1. What Washa is
1.1 In a nutshell
Washa (話者) is an app that records "who spoke, when, where, and what" from audio or video recordings. Record a meeting, interview, class or discussion, or import an existing recording file, and it does the following entirely on the device:
- Transcription (turning spoken words into text)
- Voice recognition (whose voice this is)
- Speaker separation (who said each sentence)
In the app, one recording and its record are called an "episode". The finished transcript shows each line with its speaker and time. Once you name a speaker, Washa remembers that voice and, from the next recording on, suggests "Is this this person?".
If you log in, you can also use AI proofreading, which turns the transcript into an easy-to-read "card" (Chapter 12). Recording, transcription, speakers and export all work without logging in.

1.2 Key features
| Feature | Details |
|---|---|
| Processing on the device | Transcription, voice recognition and speaker separation run on the device you use (iPhone, Mac, Android, Windows). It works without an internet connection. |
| Tells up to 10 people apart | Up to 10 speakers can be told apart in one recording. |
| Japanese mixed with English is fine | Choose "Auto (Japanese with some English)" and conversations in Japanese with some English mixed in are transcribed as spoken. |
| Many languages | iPhone and Mac support 10 languages: Japanese, English, Chinese (Mandarin), Cantonese, Korean, German, French, Spanish, Portuguese (Brazil) and Italian. Android and Windows support 5: Japanese, English, Chinese, Cantonese and Korean. Apart from the Japanese–English combination, each recording is treated as one language. |
| Free at its core, unlimited | Recording, separating speakers and using the text of who said what are free. There is no limit on the length of a recording, on monthly use, or on processing time (Chapter 3). |
| Gets smarter with a word from you | Telling Washa "how many people were talking" or "whose voice this is" improves the result. Voices that a person has confirmed stay in the Speaker library and help with the next recording. |
| Many export formats | Export as text, Markdown, Word, or captions + audio (Section 13.2). |
| 4 OSes | The same ideas and the same operations on iPhone, Mac, Android and Windows. |
1.3 About accuracy
Washa does its basic processing only on the device, without cloud AI services. So it will naturally mishear words in transcription or assign lines to the wrong speaker at times. Even so, we believe it has reached a level that is fine for practical use.
You can fix mistakes in the app:
- If the number of speakers is wrong, regroup with "Change number of speakers" (Section 10.5).
- If a line has the wrong speaker, you can reassign just that line to another speaker (Section 10.2).
- Text errors can be fixed with "Edit" → "Edit text" (Section 9.5). When the same error appears many times, use "Find and replace" (Section 9.9).
If you want the transcript tidied up, including technical terms and proper nouns, log in and use AI proofreading (Chapter 12).
2. How Washa works and where it is heading (the public scope)
This chapter covers only what the provider has made public. Details of how it works that are not written here are not public.
2.1 Washa is "the ears of AI agents"
Washa looks like a transcription app, but its real role is to be the "ears" through which AI agents perceive this world. This is the most important part of the Washa concept.
Teleport Inc. is developing Teleport's AI platform. On it live AI agents called Telepotch, which connect with various services.
Teleport's platform aims for a world like this:
- Many humans and many AI agents work together as a team.
- Not limited to one place. It runs on many platforms and devices — LINE, Slack, Discord, Teleport's own communication platform to come, and even places where you can talk in the car, like CarPlay (multi-agent, multi-platform, multi-device).
- Wherever you talk to it, an agent behaves as "one personality", and can play as a team with many people, not just on its own.
For that, agents need ears too. Washa is what we built to be those ears.
Washa has its own AI agent (a Telepotch) called "Washa-kun". In features to come, the idea is that Washa-kun understands the content and shapes it into minutes, presentations and more (this also appears on the introduction screens of the first-run guide; Section 5.2).
2.2 The engine "WashaKit" (話者キット)
Washa contains an engine developed by Teleport called WashaKit (話者キット).
- WashaKit is the engine that performs transcription, voiceprint analysis and speaker separation.
- It supports 5 OSes: iOS, Mac, Android, Windows and Linux.
- The Washa app supports the 4 of these other than Linux.
- Teleport intends to offer WashaKit in many places and keep growing it. For example, in medical settings or inside companies, it will enable uses such as transcribing huge amounts of audio in a closed environment, without sending any data to AI companies.
2.3 Recording location, and sounds other than human voices
- Washa can tell where you were when recording. Using the device's location (GPS, etc.), it attaches the place where recording started to the recording (only if you allow location; Section 7.6).
- In the future, we will make it recognize voices other than human ones too — natural sounds such as birdsong, insects and animal calls. We are growing Washa as the ears through which AI agents come to perceive this very world.
- To that end, the WashaKit engine includes a mechanism that can be extended to models that analyze sounds other than human voices.
Recognizing sounds other than human voices is a future plan. The current beta app does not have this feature.
2.4 The document format "Washa Document" (話者ドキュメント)
- Washa has a document format called Washa Document (話者ドキュメント).
- Teleport would like to offer this format as open source.
- A Washa Document is a format for laying out in time order what was happening at the scene, so it can be analyzed later. What it lays out includes not only audio but also smartphone sensor information and the various data that can be recognized from it.
In the current app, "Export to This Device" under "Help improve Washa" at the very bottom of the Export tab creates a Washa Document for one recording (one file with the transcript, the record of speaker corrections and more, and one audio file) (Section 13.3).
2.5 What this chapter does not cover
The following are not public, so they are not written in this document:
- The details of the AI models used inside WashaKit, and the processing steps
- Accuracy figures (what percentage is correct, etc.)
- When Washa-kun, Telepotch and Teleport's AI platform will be offered, and their pricing
- Terms for offering WashaKit to companies and medical institutions
- When the Washa Document format will be open-sourced, and the details of its specification
3. Pricing approach
3.1 Free at its core — and unlimited
Washa is basically a free application. This is not just for the beta period.
Recording, separating speakers, and using the text of who is speaking and how are free and unlimited.
- There is no monthly fee.
- There is no limit on the length of a recording, on monthly use, or on processing time (within what the device can handle). Processing runs on the device, so you are never charged by usage.
- Up to 10 people can be told apart in one recording.
- Recording, transcription, speakers and export work without logging in.
Because processing happens on the device, there is also the safety of the recording's contents not leaving it.
Fully usable even for free — that is Washa's biggest point.
3.2 Server features (used after logging in)
On top of that, features that use Washa's servers are available after logging in. These are in the current version:
| Feature | Details | OS |
|---|---|---|
| AI proofreading | Proofreads the transcript text on Washa's server and turns it into an easy-to-read "card" (Chapter 12) | 4 OSes |
| Card workshop (test) | Makes translation, slide, slideshow and manga cards from the transcript (Section 12.6) | iPhone, Mac |
| myWasha | Copies transcripts and cards to myWasha and shares them with read-only links (Chapter 11) | iPhone, Mac, Android |
- Usage appears as "credits" in the list of cards you made and AI-proofread versions.
- The speaker-count sheet (Section 9.2) shows a line with your plan and what is left. When not enough is left, it shows "Not enough time left", and you can choose "Plans and campaign codes ↗" or "Proofread on (date)".
- Plans, credits, payments and receipts are opened from "Manage in myWasha ↗" on the membership panel (Section 11.2).
We will keep adding optional features step by step. Examples we plan are making minutes, making infographics, and working with Washa-kun. When and what they will be is not decided yet.
4. System requirements and how to get the app
Version checked: app build 59 (0.2.0 on iPhone and Mac, 1.0.0 on Android and Windows).
4.1 Overview
| iPhone | Mac | Android | Windows | |
|---|---|---|---|---|
| Required OS | iOS 26.0 or later | macOS 26.0 or later | Android 13 or later | Windows 11 (64-bit) |
| CPU | — | Apple silicon (M series) only. Intel Macs are not supported | 64-bit ARM (arm64) only | x64 only. ARM versions of Windows are not supported |
| How to get the beta | TestFlight | TestFlight | Install the APK file directly | Installer (MSI) |
| How new versions arrive | Through TestFlight | Through TestFlight | The app updates itself | The app updates itself |
| Main devices tested | iPhone | Apple silicon Mac | Pixel 8a | Windows 11 PC |
| Internet | Needed the first time, to download the data used for processing. Not needed after that for recording and processing (features used after logging in need it) | Same as left | Needed the first time, to download the data used for processing (downloads only on Wi-Fi; you can also choose to continue on mobile data) | The processing data is bundled with the app. Normally no internet is needed for processing (it downloads at startup only when a component is missing) |
| Screen language | Japanese, English | Same as left | Same as left | Same as left |
Notes:
- iPad: You can also install the iPhone version on an iPad. However, the screens are built the same as on iPhone; there are no screens made especially for iPad. The beta is mainly tested on iPhone.
- Windows 10: Not tested. Windows 11 is the supported version.
- Android devices: It cannot be installed on 32-bit devices, x86 devices (emulators, etc.) or Chromebooks.
- Memory: No minimum memory requirement has been set. When processing a long recording, processing may stop if memory runs short (Section 9.13).
4.2 Required permissions
| Permission | Purpose | iPhone | Mac | Android | Windows |
|---|---|---|---|---|---|
| Microphone | Recording | Required | Required | Required | Must be allowed in Windows Settings (below) |
| Location (approximate) | Records the place where recording started, once | Optional | Optional | Optional | Not used |
| Speech recognition | On-device transcription | Required (asked at the import/processing entry point) | Not asked | Not asked | Not asked |
| Notifications | Alerts when processing finishes or fails (on Android, also the recording and processing indicators) | Optional | Optional | Required (recommended) | Not used |
| Camera | Speaker card photos, and photos and videos while recording | Optional | Optional (speaker card photos) | Optional | — |
| Add to Photos | Also save what you took while recording to the Photos app | Optional (only when turned on in Settings) | — | — | — |
Explanations shown on the permission prompts (iPhone, Mac):
| Permission | Japanese | English |
|---|---|---|
| Microphone | 録音と、録音の取り込みに使います。 | Used to record audio and to import recordings. |
| Location | 録音を始めた場所(おおよその位置)を 1 回だけ取り、録音に添えます。 | Gets your approximate location once when you start recording and adds it to the recording. |
| Speech recognition | 録音を端末の中で文字に起こすために使います。 | Used to transcribe recordings on this device. |
| Camera | 話者カードの写真と、録音中の写真・動画を撮るために使います。 | Used to take photos for speaker cards and photos and videos while recording. |
| Add to Photos | 録音中に撮った写真と動画を、写真アプリにも保存するために使います(設定で入にした時だけ)。 | Used to also save photos and videos taken while recording to the Photos app (only when turned on in Settings). |
- In the first-run guide, before the OS permission prompts, the "Washa uses the microphone and location" screen explains why (Section 5.3).
- You can record without allowing location. The recording simply has no place attached.
- Photos are chosen with the OS photo picker, so no permission to read the photo library is needed.
About the microphone on Windows (important)
Windows has no screen that asks for microphone permission per app. If "Let desktop apps access your microphone" is off under Windows "Settings → Privacy & security → Microphone", Washa records silence without showing an error. If no text appears even though you recorded, check this setting first.
4.3 Getting and installing the app
iPhone and Mac (TestFlight)
- From the invitation email or link, install Apple's TestFlight app.
- Install "Washa" from TestFlight.
- The "Washa" icon appears on the Home Screen (in Applications on Mac).
- New versions arrive through TestFlight. On Mac, "Check for Updates" in the menu also opens TestFlight.
- TestFlight versions may change features or display without notice (Terms of Use, Article 3).
Android (APK)
- Save the APK file you were given to the device.
- Open the file and install it. If the browser or file app asks you to "Allow from this source", allow it (this is standard Android behavior).
- "Washa" appears in the launcher.
- The first time, the app downloads the data used for processing ("Getting Washa ready" in the first-run guide; Section 5.2). It downloads only while connected to Wi-Fi. When Wi-Fi is on but not connected to a network, a message says Wi-Fi is on but there is no network connection, and you can choose "Continue on mobile data (size)".
- On a device where the processing data is incomplete, it shows "This device doesn't have the data needed to separate speakers yet, so the recording can't be processed." If this keeps appearing, please report it on the Washa Discord (Chapter 15).
- It updates itself to new versions (Section 4.4).
Windows (installer)
- The Windows version is distributed as an installer (MSI). Opening the MSI overwrites the installed Washa.
- Washa installed from the MSI updates itself from the next version on (Section 4.4).
- On first launch, preparing the bundled processing data takes about 1–2 minutes.
- If Washa is already running, a second one cannot be opened. It shows "Washa is already open. Switch to its window."
- Data is stored in the per-user location in Windows (local app data). You can change the location in Settings (Section 13.5).
Where to get the APK and MSI is announced with each distribution.
4.4 Updating to new versions (Android, Windows)
Washa on Android and Windows checks for a new version at launch, and every 3 hours while open, and updates itself.
- "Updates" in Settings shows the current status.
| Status | Display |
|---|---|
| Up to date | You're up to date (build N) |
| Downloading | Downloading build N |
| Ready (Android) | Build N is ready. It will be installed when Washa isn't recording or processing. |
| Ready (Windows) | Build N is ready. It will be installed the next time you open Washa. |
| Permission needed (Android) | To install updates automatically, allow Washa to install apps. ("Allow") |
| Couldn't install | Build N couldn't be installed |
- "Check now" checks on the spot. On Windows you can also check from the menu "Help → Check for updates…".
- On Android, allow Washa once as an app that can install from this source ("Allow from this source").
- If it shows "This is a development build. It doesn't update automatically.", it may not be a version distributed for the beta. Please let us know on the Washa Discord.
4.5 Beta expiry
- This beta works until December 31, 2026.
- From January 1, 2027, launching the app shows "The beta test period has ended. You can't make new recordings.", and new recordings and imports are no longer possible. We plan to deliver a new version before then.
- The expiry also appears in "Release notes" in Settings as "This beta works until (date)." (Section 13.5).

4.6 How the app's name is written
- It is called "話者" in Japanese, and "Washa" in Latin letters (capital W).
- "Washa" appears under the app icon and in the on-screen logo. Under the logo, "by Teleport" and the version line appear.
5. First launch
Version checked: app build 59.
On first launch, you go through the following in order:
- Agreeing to the Terms of Use and Privacy Policy (Section 5.1)
- An introduction to Washa, and preparing the data used for processing (Section 5.2)
- Your speaker card (Section 5.3)
- Trying it out (Section 5.4)
- This guide is not shown while recording. Even when you start recording from a widget and the app opens, it appears at the next launch after the recording ends.
- The introduction and the speaker card can be opened again later from Settings (Section 13.5).
5.1 Agreeing to the Terms of Use and Privacy Policy
On first launch, an agreement panel appears full-screen. The app itself cannot be used until you agree.
| Part | Japanese | English |
|---|---|---|
| Title | Washa をお使いになる前に | Before you use Washa |
| Description | 利用規約とプライバシーポリシーをお読みのうえ、同意してください。 | Please read the Terms of Use and the Privacy Policy, and agree to them. |
| Language switch | 日本語 / English | Same as left |
| Document switch | 利用規約 / プライバシーポリシー | Terms of Use / Privacy Policy |
| Checkbox | 利用規約とプライバシーポリシーに同意する | I agree to the Terms of Use and the Privacy Policy |
| Button | 同意して始める | Agree and start |
| When the document can't be read | 文書を読み込めませんでした。 | Could not load the document. |
- You can scroll the body to read the full text.
- "Agree and start" cannot be pressed until the box is checked.
- The panel's language can be switched separately from the screen language setting. Initially it follows the screen language setting (Section 13.5). With "Match system", it appears in Japanese if the device's primary language is Japanese, and in English otherwise.
- The panel has no close button, and Android's "Back" does not close it either.
- When a new version of the Terms of Use or Privacy Policy comes out, this panel appears again the next time you open the app. The current version is the 2026-10-21 edition. It appears again even for people who agreed to the previous version.
- The version agreed to and the date/time can be checked later under "About" in Settings.

5.2 Introduction to Washa, and preparing the data
After you agree, the introduction to Washa begins. While you read it, the data used for processing is downloaded in the background.
Introduction
- In the beta there are 6 pages. Each page shows "More", "Next" and "Skip", plus dots for the page number ("n of 6"). There is also "Don't show this again".
- Example titles: "Core features are free, with no limits", "Everything stays on this device", "4 core features, ready now", "Coming soon: just talk, and it takes shape", "We grow Washa together", "To our beta testers".
- The last page has "Add your voice", which takes you to your speaker card.
- On Mac and on wide-screen Android and Windows, the picture is shown on the left and the explanation on the right.
Preparing the data
| Part | Details |
|---|---|
| Title | Getting Washa ready |
| Description | A note that it is getting the data used for transcription and telling speakers apart. |
| List | For each piece of data, "Done" or "Waiting" and its size ("about N"). Sources are "From the Washa download site", "From Apple" and "From Hugging Face" |
| Total and remaining | "Total N" and "N left" |
| When finished | Ready |
- When the connection drops, it shows "Couldn't connect. Trying again in N seconds (n of 3)." and retries on its own.
- When not connected to the internet, it shows "Washa will continue when you're back online."; on Android while waiting for Wi-Fi, "Washa will continue on Wi-Fi."
- When the device is short of space, it shows "Not enough free space. Free up N more."
- When it couldn't get the data, it shows "Couldn't get Washa ready" with the reason, and "Try again".
- On devices that show "Chinese, English, Japanese, Korean and Cantonese are built in." (Android, Windows), the data for these languages is included in the app.




5.3 Your speaker card
Once Washa has learned your voice, you appear under your own name from the next recording on. It takes about a minute. The recorded voice is stored only on this device.
- Permission: The "Washa uses the microphone and location" screen explains why — "Microphone: to record your voice and your meetings" and "Location: to note where you recorded. Optional" — and "Allow" takes you to the OS permission prompts. If you decline the microphone, it shows "Washa can't use the microphone. Allow it in Settings." and "Open Settings".
- Reading aloud: "Read the text aloud" (in your usual way of speaking; slowly is fine). You read 2 short sentences ("1 of 2", "2 of 2"). Press "Start recording" and reading starts 3 seconds later.
- Check: How well it was recorded is shown.
| Result | Display (meaning) |
|---|---|
| Good | Recorded well |
| No sound | No sound was recorded. Check the microphone and record again. |
| Voice too quiet | Your voice seems quiet. Move a little closer to the microphone and record again. |
| Distorted | The sound is distorted. Move a little away from the microphone and record again. |
| Too short | Your voice was too short. Please speak a little longer. |
| Noisy surroundings | It seems noisy around you. Record again somewhere quiet. |
Continue with "Play back", "Record again" or "Next". For a minor problem you can also choose "Use it anyway". 4. Learning your voice: "Learning your voice" appears. 5. Name and photo: "Name and photo" (shown on your speaker card; you can change it later). When you are logged in, the name field shows "Filled in from your account". You can create the card without a photo ("No photo"). Create it with "Create card".
- You can put it off with "Later" along the way. If you choose "Don't set up", it shows "We won't ask again. You can create it any time in Settings › Speaker card."
- If the Speaker library couldn't be loaded, press "Start a new library" to continue. The old library is kept under a new name.




5.4 Trying it out
Once your speaker card is made, you try whether Washa can tell your voice.
- "Let's try it" (say anything; speak for about 5 seconds). Example things to say appear (such as "How is the weather today?").
- When it recognizes you, it asks "Is this (name)?". "Yes" finishes.
- On "No", it shows "So that was someone else. You can speak again or finish.", and you can choose "Speak again", "Finish" or "Record my voice again". The speaker card you made is kept as is.
- Finally, "Ready" and "From your next recording, you'll appear as "(name)"." appear, and "Get started" begins.


5.5 Storage location (Mac, Windows)
- Mac and Windows keep recordings and speaker cards in a default location. You can change it with "Where to keep your files" in Settings (Section 13.5).
- Changing the location does not move the contents automatically. If you want to move them, move the folder yourself.
- If the storage location can't be found, it shows "Washa can't find your storage folder. Connect the external drive, or choose the location again."
6. Overview of the screens
Version checked: app build 59.
6.1 Common screen structure
| Screen | Role |
|---|---|
| Home | Shelves of recent recordings, Continue, your cards, news and suggestions |
| Episodes | The list of recordings, by date |
| Speakers | Cards for each person in the Speaker library |
| Transcript screen | Read, listen to and correct the transcript or card of one recording |
| Settings | Account, display, language, camera, background, logs, Trash, About, and more |
In addition, "sheets" (sheets that rise from the bottom of the screen, or panels in the middle of the window) appear on top when needed: the speaker-count sheet, the processing sheet, the speaker panel, the share and export sheet, and so on.
6.2 The bottom bar (iPhone, Android)
At the bottom of the screen on iPhone and Android is this bar:
| Position | Items |
|---|---|
| From the left | Home / Episodes / ● REC / Speakers / Settings |
| Upper left of REC | A small round button for the transcription language |
| Upper right of REC | A small round button for the microphone input (when there are 2 or more microphones) |
- The big "REC" in the middle starts recording (Chapter 7).
- When you are logged in, "Settings" shows your face and level ("Settings · level").
- When you play a recording, a media bar floats above the bar. Press it for full-screen playback; playback continues even if you close the Transcript screen.

6.3 Home
Home stacks these shelves from the top:
| Shelf | Contents |
|---|---|
| Recent | Recent recordings |
| Continue | Recordings you played partway ("N min left") |
| Your cards | Cards made by AI proofreading or the card workshop |
| News | News sent from the server |
| Suggestions | Suggestions sent from the server |
- News and suggestions come from the server. When offline, the last ones received are shown.
- Shelf tags let you filter ("All", "My cards", etc.).
- Press a card to open it.
- When there is nothing, "Nothing here yet" and "Recent recordings and where you left off will show up here after you record." appear.
- On Mac, choosing "Home" in the left column shows the shelves in the right pane.


6.4 Episodes (the list of recordings)
The top part
- The logo "Washa" and "by Teleport", with the version line below (e.g. "0.2.0 · build 59")
- Filter tags "All / Favorites". "Favorites" shows only recordings with a heart (when there are none: "No favorites yet" and "Open a recording and add a heart to collect it here").
- The sort icon ("Sort"): "Date Recorded" (the usual order), "Date Imported" (the order they were added to Washa) and "Date Edited" (the order they were last edited). When not in the usual order, the icon is colored. The chosen order is kept after closing the app.
- "+" (import a recording; Chapter 8)
- Recordings that are processing, queued, or couldn't be processed always appear at the top of the list. Up to 3 are listed one by one; from the 4th on there is one line, "N more". A processing row shows "%" and a line for how far it has got.
Date headings
- "9月24日(水)" ("Wed, Sep 24" in English) and the number of recordings that day, "N recordings". Where a day breaks depends on the date used by the chosen sort order.
Recording rows
- Time on the left (e.g. "14:00")
- Title (one line, to the right edge). A heart mark if it is a favorite
- An unread dot for recordings not opened yet
- A small line of text: place name, number of people ("N people"), length ("N min"; recordings under 30 seconds show "Under 1 min"). When the recording date is unknown, "Date imported"; if it was imported on a different day, "Imported 9/24". The length is shown right after importing and right after recording ends.
- On the right, the faces of the speakers in it (a photo, a picture or a colored circle; unnamed speakers show "?")
- Only when processing did not finish normally, a status tag (Section 9.1)
Row actions
| Action | iPhone | Mac | Android | Windows |
|---|---|---|---|---|
| Open | Tap | Click | Tap | Click |
| Favorite | Swipe left to right | Right-click, "…" | Swipe right | Right-click, "…" |
| Menu | Swipe right to left, long-press, "…" | Right-click, "…" | Long-press, "…" | Right-click, "…" |
| Delete (key) | — | Delete key | — | Delete key |
Main items in the row menu: "Add heart / Remove heart", "Rename", "Mark as unread", "Start Over…", "Transcribe Again from Audio…", "Export Audio", "Get Info", "Export logs" and "Delete". For a recording that is processing, "Stop processing" appears instead of "Delete". Which items appear, and in what order, depends on the recording's state and the OS.
- "Mark as unread": puts the unread mark back on a recording you have read.
- "Start Over…" and "Transcribe Again from Audio…": Section 9.12.
- "Export Audio": you can share or save the audio file as is, even for a recording that is mid-processing or failed.

6.5 Speakers
A list of each person whose name is confirmed (speakers in the Speaker library). The filter tags are "All / Favorites".
- At the top, the row "Speaker library (N)". Press it to open the whole Speaker library (Section 10.7).
- Cards for confirmed people: photo or picture, name, and "N recordings · M lines". Press a card to open the list of recordings that person appears in. On Mac and Windows, the right pane shows the speaker card next to that person's recordings (when the right pane is narrow, only the list is shown; open the card from "Open speaker card" at the top).
- The entry for "N speakers without a name yet": a list of recordings with unnamed speakers.
- When there is nothing: "No speakers yet".

6.6 Screen differences by OS
| iPhone | Mac | Android | Windows | |
|---|---|---|---|---|
| Screen layout | Switch with the bottom bar; screens stack one at a time | 3 columns (left column, list in the middle, Transcript screen on the right) | Switch with the bottom bar. Columns on wide tablets | 3 columns (same build as Mac) |
| Left column | — | Home / Episodes / Speakers / Recently Deleted (⌘1–⌘4). At the foot, the record button and the account row | — | Same build as Mac |
| How to open Settings | "Settings" in the bottom bar | Menu "Washa → Settings…" (⌘,), separate window | "Settings" in the bottom bar | Menu "File → Settings…" (Ctrl+,), separate window |
| Menu bar | None | Yes (Section 6.7) | None | Yes (Section 6.7) |
| How the speaker panel appears | A sheet from the bottom | The inspector to the right of the Transcript screen | A sheet from the bottom | A panel inside the window |
| Version line | Under the logo | Under the logo at the top left of the window | Under the logo | Under the logo |
- On Mac and Windows, no sheet appears when you stop recording or add files. The right pane shows the progress, time left, the number queued and "Stop processing". The speaker-count question and confirmations for deletions that can't be undone appear as sheets.
- On Mac and Windows, only one recording bar appears, at the top of the right pane. Notices appear over the top of the right pane, and the playback bar is at the bottom of the right pane.
- On Mac, ⌘F searches titles, places and speaker names. The Transcript screen's title becomes the window title, and you can rename from there.
- When nothing is selected, the right side shows "Select a recording in the list on the left to read it here."


6.7 Mac and Windows menus and keyboard operations
Mac
| Menu | Item | Key |
|---|---|---|
| Washa | Check for Updates (opens TestFlight), License | — |
| File | New Recording | ⌘N |
| File | Import Recording… | ⌘O |
| File | Export… | ⌘E |
| File | Get Info | ⌘I |
| File | Export Audio | — |
| Edit | Bookmark (add or remove on the line now playing) | ⌘D |
| View | Home / Episodes / Speakers / Recently Deleted | ⌘1 / ⌘2 / ⌘3 / ⌘4 |
| View | Larger Text / Smaller Text / Default Size | ⌘+ / ⌘− / ⌘0 |
| Recording | Add Bookmark | ⌘B |
| Playback | Play/Pause / Back 10 Seconds / Forward 10 Seconds | Space / ← / → |
- The key for a new recording changed from ⌘R to ⌘N.
- Settings opens with ⌘,. ⌘F searches.
- A recording selected in the list can be deleted with the Delete key (a confirmation appears).
- Panels close with Esc, and the main button of a confirmation panel can be pressed with Return.
- In the speaker identification flow, ⌘Z undoes.
Windows
| Menu | Item | Key |
|---|---|---|
| File | New recording / Import recording… / Export… / Get info / Settings… / Exit | Ctrl+R / Ctrl+O / Ctrl+E / Ctrl+I / Ctrl+, / — |
| Edit | Cut / Copy / Paste / Select all (when in a text field), Bookmark | Ctrl+X / Ctrl+C / Ctrl+V / Ctrl+A, Ctrl+D |
| View | Text size (choose from 5 steps), Larger text / Smaller text / Default size | — / Ctrl++ / Ctrl+- / Ctrl+0 |
| Playback | Play/Pause / Back 10 seconds / Forward 10 seconds | Space / ← / → |
| Help | Check for updates… / License | — |
- Esc closes panels one at a time, starting from the top one (see also known issue K-16).
- The Delete key brings up the confirmation to delete the open recording.
- The text size setting applies to the whole screen.
7. Recording
Version checked: app build 59.
7.1 How to start recording
| OS | Entry points |
|---|---|
| iPhone | "REC" in the middle of the bottom bar / "Record with Washa" in Control Center and on the Action button / Home Screen and Lock Screen widgets / "Start recording with Washa" in the Shortcuts app |
| Mac | The record button at the foot of the left column / "Record" in the toolbar / menu "File → New Recording" (⌘N) |
| Android | "REC" in the middle of the bottom bar / the Quick Settings tile "Record with Washa" / the Home screen widget / "Continue recording" in the notification shown when recording stopped |
| Windows | "● Record" / menu "File → New recording" (Ctrl+R) |
- The first time you press it, it asks for microphone permission, then location permission (except on Windows). If you already allowed them in the first-run guide, it does not ask.
- If you don't allow the microphone, it shows "Washa needs microphone access to record. Nothing has been recorded yet." and "Open Settings".
- On Windows, when no microphone is found, it shows "No microphone was found on this PC. Connect a microphone and press again. Nothing has been recorded yet."
- "REC" on the widget starts recording on the spot without opening the app (Washa opens only when microphone permission hasn't been given yet, to ask for it; Section 7.8).
- If, before recording starts, the device doesn't have enough free space for an hour of recording, recordings in the Trash are cleared, oldest first, to make room.
7.2 The recording screen
| Part | Details |
|---|---|
| Title | "Recording" / "Paused" / "Microphone unavailable. Waiting…" / "Connecting to the microphone…" |
| Elapsed time | e.g. "0:00:04" |
| Level bar | Lets you see whether voice is coming in |
| Spoken language | A small button. Press it to choose the language (Section 7.4) |
| Input tag | "Input: This iPhone's microphone", etc. Press it to choose the microphone (Section 7.5) |
| Interruption count | "Interrupted N times". Press it for the "Interruptions" list (Section 7.7) |
| Remaining capacity | Only when the device's free space drops below 30 minutes' worth, "N min of recording left" |
| Camera | On iPhone and Android, when the setting is on. You can take photos and videos (Section 7.10) |
| Buttons | "Pause" and "Stop". While paused, "Resume" and "Stop" |
| Live transcript | Text flows in as you speak (Section 7.3). There are copy and share buttons |
How it appears on each OS:
-
iPhone: The recording panel appears at half height. Pull it up to enlarge it, and further to full screen. Closing the panel keeps recording, and a red bar appears on screen. Press the bar to go back to the panel.
-
Mac and Windows: Not a panel; one recording bar appears at the top of the right pane. The recording screen is given plenty of space.
-
Android: Press the bar to open the recording panel (it opens at half height; pull it up for full screen).
-
The device vibrates briefly when recording starts and stops, and so on. If recording stops unintentionally, a longer vibration lets you know (iPhone, Android).
-
If there is no speech and no text for 5 minutes, a notification says "Keep recording?" and "No one has spoken for N minutes." "Keep recording" waits another 30 minutes. If you do nothing, it doesn't stop (it is only a notification).
-
Press "Stop" and "Saving recording…" (progress bar and %) appears; the recording goes into the list, and then the speaker-count sheet appears (Section 9.2).



The recording is first named automatically, like "Recording Oct 7, 1:05 PM" ("10月7日 13:05 の録音" in Japanese). You can change it later (Section 9.7).
7.3 Live text during recording (provisional text)
While recording, what is said flows onto the screen as text.
- This is provisional text, and speakers are not separated. After recording ends, processing creates the transcript with speakers separated (the clean copy).
- On iPhone and Mac, confirmed text is dark gray, and text that may still change is light gray.
- "Copy transcript so far" and "Share transcript so far" let you take the text out even while recording.
- Until processing finishes, it stays on the Transcript screen as "Live transcript (draft)".
- Even if other recordings are being processed while you record, the live text does not stop.
Main messages when no text appears:
| Situation | Display (meaning) |
|---|---|
| Preparing | Preparing live transcript… |
| Showing (iPhone, Mac) | This is provisional text; it is cleaned up after recording. |
| Low Power Mode | Because of Low Power Mode, only finished sentences are shown. |
| Device is hot | The device is hot, so the text is paused. Recording continues. |
| Device not supported | Live text isn't available on this device. It is transcribed after recording. |
| Language not supported | Live text isn't available for this language. It is transcribed after recording. |
| Stopped | Live text has stopped. Recording continues. |
In every case, the recording itself continues. Even without live text, the recording becomes a transcript in processing after recording.
When no sound is coming in during recording, it tells you with messages like "We can't hear anything. Check the microphone. Still recording." and "The sound is very quiet. Check where the microphone is. Still recording."
7.4 Choosing the spoken language
The transcription language is set in these 3 places:
- "Transcription language" in Settings: the language you usually use. The "Main language" is set automatically from the device language and can be changed later.
- The language button on the recording screen and in the bottom bar: the language of the recording you are about to make.
- The speaker-count sheet: you can set a different language just for that recording (Section 9.2).
The language chosen for the previous recording is not carried over to the next one automatically.
| Option | Japanese | Supported OSes |
|---|---|---|
| Auto (Japanese with some English) | おまかせ(日本語・英語まじり) | All OSes |
| Japanese | 日本語 | All OSes |
| English | 英語 | All OSes |
| Chinese (Mandarin) | 中国語(普通話) | All OSes |
| Cantonese | 広東語 | All OSes |
| Korean | 韓国語 | All OSes |
| German | ドイツ語 | iPhone, Mac |
| French | フランス語 | iPhone, Mac |
| Spanish | スペイン語 | iPhone, Mac |
| Portuguese (Brazil) | ポルトガル語(ブラジル) | iPhone, Mac |
| Italian | イタリア語 | iPhone, Mac |
- For conversations in Japanese with some English mixed in, choose "Auto". "Auto" listens to about the first 30 seconds, estimates the language, and tells you "Detected: (language)" and "Other candidate: (language)". If the estimate is wrong, choose again from the candidates.
- Other languages are treated as one language per recording. For example, a recording mixing Chinese and Korean cannot be transcribed correctly in both.
- Cantonese is shown in the Traditional characters used in Hong Kong.
- iPhone and Mac: When the data for the chosen language isn't on the device, a download prompt appears ("Washa will download (language) data. Wi-Fi recommended."). Language data is shared with the OS's dictation. Up to 5 languages at a time can be used for transcription; trying to download a 6th shows "You can use up to 5 languages. Remove one to download another." Choose a language to remove and press "Remove and Download", or "Cancel" to remove nothing.
- Android and Windows: Data for Japanese, English, Chinese, Cantonese and Korean is included from the start.

7.5 Choosing the microphone
- Press the input tag ("Input: …") or the microphone input button in the bottom bar to see the list of available microphones. The built-in microphone is shown as "This iPhone's microphone", "This Mac's microphone", "This device's microphone" (Android) or "This PC's default microphone" (Windows).
- When another microphone, such as Bluetooth, is available, it shows that it can also be used. Microphones that give phone-call sound quality are marked "Phone-call quality".
- The microphone chosen in Washa does not change the OS-wide default input.
- If the microphone in use is disconnected during recording, recording continues with another microphone, and "Microphone disconnected. Recording with another microphone." appears. When the OS moves the input to another microphone during recording, it shows the actual input and tells you the input has changed.
- On iPhone, when the OS "Voice Isolation" is on, a caution appears: "Voice Isolation may filter out people farther from the microphone". For meetings with several people, it is better not to use Voice Isolation.
7.6 Recording location
- If you allow location, when recording starts it gets your approximate location just once, turns it into a place name (down to city and town), and attaches it to the recording. The place name is shown in the chosen language.
- The place name appears in the list row, in the line under the title on the Transcript screen, on the recording screen, and in the first line of exports.
- You can fix the place by hand in the "Name and date" sheet (Section 9.7). Leaving it blank removes it.
- Imported files have no place attached.
- Windows does not get your location. You can only type a place by hand.
7.7 Interruptions and automatic resume
When the microphone becomes unavailable — a call comes in, earphones are unplugged, another app uses the microphone, and so on — recording does not stop but goes into a "waiting" state.
- The title changes to "Microphone unavailable. Waiting…", and when the microphone is available again it continues recording automatically. You can also resume yourself with "Resume now".
- The number of interruptions appears as "Interrupted N times"; press it to open the "Interruptions" list. Each row shows the time, the reason, and "Resumed" or "Waiting for the microphone".
- Example reasons: "Another app or a phone call used the microphone", "Microphone or headphones disconnected", "Microphone was muted".
Recordings are saved in chunks every few minutes, so even if the app crashes, what was recorded up to then remains.
- A recording that couldn't be finished, because the app crashed or similar, appears in the list the next time you open it as "Interrupted recording (N min)", with the reason and the buttons "Use this recording" and "Delete". It is not processed automatically.
- On Android, after the app crashes or the device restarts, a "Recording stopped" notification appears. Press it to continue recording.


7.8 OS-specific recording notes
iPhone
- Recording continues when the screen is off or you switch to another app.
- The recording indicator (Live Activity) appears on the Lock Screen and in the Dynamic Island, with the elapsed time, place name, interruption count, and "Pause" and "Stop" buttons. After 8 hours the Lock Screen display disappears under OS rules, but recording continues (the app shows "The Lock Screen no longer shows the recording, but recording continues.").
- You can put the "Record with Washa" control in Control Center and on the Lock Screen. Pressing it starts recording on the spot. While recording it pauses; while paused it resumes. It can also be assigned to the Action button (Settings → Action Button → Controls) and works the same way.
- From the Shortcuts app, you can use the actions "Start recording with Washa", "Pause recording", "Resume recording" and "Stop recording". There is no fixed phrase for asking Siri.
- You can put widgets on the Home Screen and Lock Screen (Section 7.9).
- Pressing "Stop" on a widget, the Lock Screen or the Dynamic Island turns it into a big full-width "Stop"; press again to stop.
- While recording, the processing indicator is taken down. It is shown again when recording ends.
Mac
- Only the microphone can be recorded. Sound playing on the computer (system audio) cannot be recorded. To record the other side of an online meeting, the microphone has to pick up the sound from the speakers.
- Closing the lid puts the Mac to sleep and stops recording ("Closing the lid puts your Mac to sleep and stops recording."). While recording, automatic sleep due to inactivity is prevented.
- There is no Lock Screen display, widget or shortcut.
Android
- Recording continues when the screen is off or you switch to another app. While recording, the notification shows "Recording" with the elapsed time, and "Pause" and "Stop" (also on the lock screen). On Android 16 and later, a recording chip appears in the status bar.
- Pressing "Stop" in the notification changes it to "Stop recording?", and you choose "Cancel" or "Stop".
- You can put a "Record with Washa" tile in Quick Settings.
- You can put a widget on the Home screen (Section 7.9).
- If notifications are not allowed, it shows "Notifications are off, so Washa can't alert you if recording stops."
- If battery usage is set to "Restricted", recording stops when you leave the screen. If you see the message saying so, change the app's battery usage in Android settings to "Optimized" or "Unrestricted" before you start.
Windows
- Closing the window during recording stops and saves the recording before closing ("Saving the recording before closing…").
- There are no notifications or displays outside the window.
7.9 Widgets (iPhone, Android)
- Recording widget: When not recording, "REC" (press it to start recording on the spot without opening the app) and "Open app". While recording, "● REC" and the elapsed time on a red background, with "Pause" and "Stop". While paused, "Paused" on a dark gray background, with "Resume" and "Stop". The large size also lists recent recordings.
- "To check" widget: Shows the number of recordings that have speakers without names yet. Press it to open "Speakers".
7.10 Camera while recording (iPhone, Android)
You can take photos and videos while recording. This appears only when "Camera while recording" is on in Settings (Section 13.5).
- The camera button on the recording screen ("Open camera (recording continues)") opens the camera. The transcript is shown at the top and the camera view at the bottom.
- Take shots with "Take a photo", "Start video" and "Stop video". Videos have no sound (the sound goes into the recording). Hold the device sideways to shoot in landscape.
- When you take one, "Photo saved (#n in this recording)" appears. What you take is attached to that moment of the recording.
- "Open album" lets you look back at what you took ("Photos and videos: N") and delete items. Deleting them leaves the recording and the transcript.
- After recording, you can also add shots with the camera on the Transcript screen ("Attaches at (time)").
- Photos you took appear as illustrations after paragraphs of the AI-proofread card. You can "Replace", "Remove" or go "Back to auto".
- In export, you can choose the RO-Crate format, which bundles the recording, the transcript and what you took (Section 13.2).
- When camera use isn't allowed, a message saying so and "Open Settings" appear. In every case, recording continues.
7.11 "Text only" recording (no speaker separation)
Choose "Text only (no speakers)" in the "+" menu to start a recording that makes the transcript only from the text shown during recording, without separating speakers.
- Since speakers are not separated, you can read it right after recording ends. The speaker-count sheet doesn't appear either.
- At the top of the Transcript screen, "Speakers not separated yet" and the button "Separate speakers" appear. Press it later to run speaker separation.
- On devices that can't show live text, it shows "Live transcript isn't available on this device. When you finish, choose "Separate speakers" to transcribe it."
8. Importing recordings
Version checked: app build 59.
8.1 Importable formats
- You can import audio and video files. For video, only the audio part is used.
- Formats confirmed to work: m4a, mp3, wav. On Windows, the audio in mp4 can also be read.
- When a format can't be read, it shows "Couldn't read this audio format. Supported formats are m4a, mp3, and wav." and "Choose another file".
- If you mix in files that are neither audio nor video, it shows "Files that aren't audio or video can't be imported, so they were skipped."
- You can import any number at once. You can add more even while other recordings are being processed (they join the queue and are processed automatically in turn).
8.2 Entry points by OS
iPhone
- "+" in Episodes → "From Files": choose from the Files app.
- "+" → "From Voice Memos": a guide sheet appears.
- Title "Import from Voice Memos", description "To import a voice memo, send it to Washa from the Voice Memos app."
- Steps: "1. In Voice Memos, select the recording you want to import", "2. Choose Washa from Share", "3. Importing starts, and the recording appears in the list when done"
- Buttons "Open Voice Memos" and "Close"
- Share button in other apps → "Washa": the "Send to Washa" screen appears (1–20 items at a time; audio and video only). Press "Send" and it shows that N items were sent and that importing starts when you open Washa. Importing starts when you open Washa.
Mac
- "+" → "From Files", or menu "File → Import Recording…" (⌘O).
- Drag and drop files onto the window. While hovering, "Release to import" appears. Folders can't be imported ("Folders can't be imported. Drop audio files instead.").
- "+" → "From Voice Memos": a guide sheet (same as iPhone; step 2 adds that on Mac you can also drag the recording into this list).
- Share from Finder or Voice Memos → "Washa": Washa comes to the front the moment you send, and importing starts.
Android
- "+" → "From Files": choose with Android's file picker (multiple allowed).
- "+" → "From other apps": a guide sheet appears. Title "Import from other apps", steps "1. In your recording app, select the recording you want to import", "2. Choose Washa from Share", "3. Importing starts, and the recording appears in the list when done".
- Share from other apps → "Washa".
Windows
- "+" → "From Files", or menu "File → Import recording…" (Ctrl+O).
- Drag and drop files onto the window ("Release to import").
- "+" → "From other apps": a guide to the drag-and-drop steps appears.
- You can also import by dropping a file on Washa.exe, or with "Open with" in File Explorer.

8.3 Recording date
By default, the list of recordings is ordered by "date recorded" (for sorting, see Section 6.4). For the date recorded, the first of the following that is known is used:
- The time recorded in Washa
- A date fixed by hand
- The date in a Washa Document
- The creation date of the original file
- The creation date written inside the file
- The date imported (in this case the list row adds "Date imported")
- The date is not guessed from the file name.
- If the import was on a different day from the recording, the row adds something like "Imported 9/24".
- The date recorded can be fixed in the "Name and date" sheet (Section 9.7).
9. Processing, and reading and correcting the transcript
Version checked: app build 59.
9.1 The processing flow and status tags in the list
When you finish recording or import, things proceed in this order:
- The speaker-count sheet appears (Section 9.2). You can answer later.
- Processing (5 stages; Section 9.3). However many recordings there are, they are processed automatically in turn.
- It becomes readable (read the transcript on the Transcript screen; Section 9.4). For recordings where you answered the number of speakers, speakers are regrouped into that number.
- For recordings where you are logged in and turned on "Proofread with AI automatically", AI proofreading follows (Chapter 12).
Processing goes one recording at a time, in order. While you are recording, processing pauses to give priority to recording, and continues from where it was once recording ends.
The list row shows a status tag only when things did not finish normally.
| Status | Tag |
|---|---|
| Queued | Queued |
| Processing | The stage name and progress (e.g. "40% · About 3 min left") |
| Not processed | Not processed yet (a recording whose processing was stopped before it began) |
| Processing interrupted | Processing interrupted — (reason, e.g. because you switched to another app) |
| Processing stopped | Processing stopped — you can resume |
| Couldn't process | Couldn't process. Open it to see the reason and how to retry |
| No lines | No lines |
When you open a queued recording, one sentence tells you why it is waiting.
| Reason for waiting | Display (meaning) | Main button |
|---|---|---|
| Another recording is being processed | Starts when the other processing finishes (waits while recording). | Process next (puts it next without stopping the recording being processed) |
| Waiting for the recording to end | Starts when recording ends. | — |
| Tidying up the Speaker library | Tidying up the Speaker library; starts when that is done. | Start now |
| Waiting for transcription permission | Waiting for transcription permission. | Allow and start |
| Waiting for Washa to be opened | Starts when you open Washa. | — |
9.2 The speaker-count sheet
Appears after you finish recording or import (not for "Text only" recordings). However many you import at once, there is one sheet.
| Part | Details |
|---|---|
| Name tag | The recording's title and "8 min 31 s · place name". When several were imported, "N recordings" |
| Question | How many people were talking? |
| Number | 1–5 and "More ▾" (6–10). "Not sure" leaves it to the app |
| Spoken language | Shown on one line. Press it for the sheet to choose (Section 7.4) |
| Playback bar | Lets you listen to the recording to check |
| AI proofreading | "Proofread with AI automatically" on/off and its setting ("Standard" or "Detailed"). Below it, a line with your plan and what is left |
| Trash | Top right. Lets you delete an unneeded recording on the spot |
| Lead line | Processing goes in order. Choosing the number and language makes Washa correct it accordingly |
| Buttons | "Apply". It can't be pressed until you choose the number; it shows "Choose the number of speakers to apply". The close button is "Later" |
- When you know the number of people, always choose it. The result gets better.
- While the sheet is shown, and even if you close it ("Later"), processing goes on in order. If you answer the number later, speakers are regrouped; if you change the language, it is processed again in that language.
- When you are not logged in, the AI proofreading row shows "Log in for free".
- When not enough AI proofreading is left, it shows "Not enough time left", and you can choose "Plans and campaign codes ↗" or "Proofread on (date)" (AI proofreading starts on the day the allowance comes back).
- On Mac and Windows too, the speaker-count question appears as a sheet.



9.3 The processing screen
Processing goes through 5 stages.
| Stage | Display (in progress) | Japanese |
|---|---|---|
| 1 | Preparing audio… | 音を整えています |
| 2 | Separating speakers… | 話者を聞き分けています |
| 3 | Transcribing… | 文字に起こしています |
| 4 | Matching lines to speakers… | 誰の発言かを合わせています |
| 5 | Finishing up… | 仕上げています |
The processing sheet (iPhone, Android) shows the following. Mac and Windows show them in the right pane instead of a sheet.
- The title "Import recording" ("Resume processing" when continuing) and the recording's name
- The stage name, a progress bar, and progress and time left such as "50% · About 3 min left". The "%" is shown even while the time left isn't known yet. Time left appears as "Almost done", "About N min left", "N–M min left" and so on. Only while needed components are being downloaded, a message says so.
- Tags for the 5 stages (how far it has got)
- If others are queued, a line saying N more recordings are waiting
- "Import more recordings" (you can add more even while processing)
- "Stop processing"
A thin bar "Transcribing · 42% · About 1 min left" with [Stop] also appears at the top of the Transcript screen. When done, it turns into the transcript on the spot.
Stopping processing
- A confirmation appears: "Processing will stop. The recording stays in your list, and you can resume processing later."
- The main button is "Keep processing". When others are queued, you can choose "Stop processing this recording" or "Stop all remaining processing too".
- A stopped recording stays in the list, and you can resume processing later.
When processing finishes
- The processing sheet shows "Transcript ready" with "Transcript" and "Back to list".
- When you are on another screen, a one-line notice that the transcription of "(title)" is done, with "Open", appears at the top or bottom of the screen.
- On iPhone, Mac and Android, a notification arrives for each recording when its transcription finishes and when it couldn't be processed. Press the notification to open that recording.


Processing when you leave the app
| OS | Behavior |
|---|---|
| iPhone | When processing could be handed to the iOS system, it continues even if you switch to another app, and progress shows on the Lock Screen ("Processing continues if you switch apps. You can see progress on the Lock Screen."). When it couldn't be handed over, a message asks you to keep this app open. Stopped processing continues automatically from a checkpoint when you open the app. While processing with the app open, the screen is kept awake so Auto-Lock doesn't stop it |
| Mac | Processing continues even if you close the processing sheet ("Processing continues even if you close this window. When it's done, the recording appears in the list.") |
| Android | The notification shows a processing indicator (recording name, stage, time left, and "Stop this one"), and processing continues in the background |
| Windows | Processing continues while the window is open |
- Even when processing stops partway, stages already done are not thrown away; it continues from there.
- The iPhone Lock Screen and Dynamic Island also show a processing indicator (stage name, %, time left, "N more", and "Stop this one").
9.4 The Transcript screen (transcript and card)
For a recording that has an AI-proofread card, the card is shown as is when you open it (Chapter 12). The original transcript (the source version) is opened from the "Version" section of "…", and "Back to the card" returns. While viewing the source version, "Viewing the source version" is shown.
The top part
- The recording's title, with a pencil next to it (fix the name and date; Section 9.7)
- Under the title, one line of "place · date and time recorded · total length" (when there is no place, only the date/time and length). Press it to open the recording's info (Section 9.10)
- Buttons in the top bar: camera (on iPhone and Android when the setting is on), bookmarks, share (the share and export sheet; Section 13.1) and "…" (Section 9.8)
- The to-do line: only when there is something to do now, one line appears at the top.
| Line shown | Button |
|---|---|
| There are n people without a name | Decide |
| You changed the number of speakers, so the card is out of date | Redo |
| You edited the source, so the card is out of date | Redo |
| (length) trimmed from the start (the same for the end) | Restore |
- "Speakers in this recording" and the row of speaker tags. Unnamed speakers have a "?". Under the tags, a box with just the one thing to do now may appear (such as "Identifying speakers (n of N) is in progress" and "Continue identifying speakers").
- "Change number of speakers" (Section 10.5)
- "Check speakers" (Section 10.2)
- A recording not yet AI-proofread shows a big "Proofread with AI". When not logged in, it is "Log in to proofread with AI" (Section 12.1).
Body
- Each line shows the speaker tag, time and text.
- Backchannel responses (short "uh-huh"s) appear as faint lines. Short backchannels whose words were not transcribed get no tag.
- Lines marked "Noise" stay in faint text (Section 10.6).
- Bookmarked lines get a mark.
- Lines whose speaker you changed by hand get a ✎ mark ("✎ marks lines whose speaker you changed manually.").
Listening
- Press a line to play from there.
- The playback bar at the bottom: back 10 seconds, play/pause, forward 10 seconds, a position bar, and playback speed (a gauge-shaped icon). On iPhone and Android it is split into 2 rows.
- During playback, the text follows the line now playing. While you scroll by hand it doesn't follow, and returns when you stop.
Line menu (each line's "…" or long-press)
- "Bookmark" ("Remove bookmark" if it already has one)
- "Change speaker"
- "Edit"
- "Play from here" (iPhone, Mac)


9.5 Correcting lines (Edit)
On the right of the panel at the bottom of the Transcript screen is the "Edit" button. Press it and the same spot changes to "Done", and "Select all" appears on the left. Press lines to select them.
- Depending on how many are selected, the actions line up in one row: "Edit text", "Split line", "Change speaker", "Merge lines" and "Delete line".
- Actions that can't be used are faded, with one line explaining why.
- Nothing selected: "Select the lines to edit"
- Non-adjacent lines selected: "Only adjacent lines can be merged"
- All lines selected: "You can't delete every line"
- The playback bar stays visible while editing. You can correct while listening.
Edit text and split line
- Edit text: fix the text in the field and press "Save".
- Split line: press the character where you want to split ("Select where in the text to split"), then "Split here". The text from that character on is colored, and a vertical divider appears.
- When merging lines from different speakers, it asks "Who said the merged line?".
Results and undo
- Right after a correction, one line appears in the panel — "Text updated.", "Line split.", "Merged N lines into one." or "Deleted N lines." — with "Undo" (it disappears after 6 seconds).
- There is no confirmation screen when deleting. It is deleted right away, so if you make a mistake, press "Undo".


9.6 Bookmarks (栞)
- Add one with "Bookmark" in the line menu. On Mac ⌘D, and on Windows Ctrl+D, adds or removes a bookmark on the line now playing. On Mac you can also add one while recording with the menu "Recording → Add Bookmark" (⌘B).
- The bookmark button at the top opens the list: the title "Bookmark" and "+ New bookmark".
- In selection mode, "Select all" appears in the row above. It bookmarks every line, including backchannels. When all are bookmarked, it becomes "Deselect all".
- Press a row in the list to play from there. Swipe a row to delete it (on Android and Windows, with the delete action).
- When empty, it shows "Add one with "+ New bookmark" above, or from the "…" menu on a line."
- Bookmarks remain after regrouping speakers. However, if line boundaries changed and some couldn't be reattached, a message tells you how many bookmarks couldn't be reattached.
- Bookmarks are not included in exported text.

9.7 Correcting the name, date and place
"Rename" in the row menu, or the pencil next to the title on the Transcript screen, opens the "Name and date" sheet. On Mac you can also rename from the window title.
| Field | Details |
|---|---|
| Recording name | The recording's title |
| Date recorded | Date and time. A future date/time can't be set ("Can't be set to a future date") |
| Location | The place name. "Where this was recorded. Leave blank to remove it." |
- Under the date field, it shows where the current date came from. For example "Time recorded in Washa", "Date you set manually", "Date from the exported document", "Creation date of the original file", "Recording date unknown, so the import date is used".
- The buttons are "Save" and "Cancel".

9.8 The "…" on the Transcript screen
The "…" on the Transcript screen groups these items:
| Item | Details |
|---|---|
| Find and replace | Section 9.9 |
| Trim start and end | Section 9.11 |
| Photos taken | The album of what you took during and after recording (iPhone, Android) |
| Get Info | Section 9.10 |
| Version | Card versions ("Card · v n · AI", "Card · v n · Edited") and "Source version" |
| Redo the card… | Section 12.4 |
- For a recording without a card, "Version" and "Redo the card…" don't appear.
- For a recording that is processing or couldn't be processed, "…" shows "Get Info" and "Export Audio".
- On Mac, "Trim start and end" is also in the "…" in the toolbar.
9.9 Find and replace
Open it with "…" → "Find and replace". When viewing the card it says "Searching the card"; when viewing the source, "Searching the source".
- Type in the "Search" field and it shows "n of N". Move with "Previous" and "Next". When nothing is found, "No Results".
- Type in the "Replace with" field, then "Replace" or "Replace all". When done, it shows how many places were replaced.
9.10 Recording info (iPhone, Mac, Android)
The line under the title, or "Get Info" in "…", opens the "Recording Info" panel. On Mac, "File → Get Info" (⌘I) also opens it.
- It lists the title, place, date/time recorded and imported, total length, audio format and size, speakers, language, the version used for processing, the number of corrections, and the receipt number if it was sent to the development team.
- "Copy Info" copies all of it as text. When reporting a problem, pasting it makes things certain.

9.11 Trim start and end (iPhone, Mac, Android)
You can leave the chat at the start of a recording, or the sounds of packing up at the end, out of the transcript.
- With "…" → "Trim start and end", move the left and right handles ("Range start", "Range end") to set the range to use. The handles snap to breaks between lines.
- The sheet shows "First line", "Last line", "n lines hidden" and "Not speaking in this range: (names)". You can check with "Play from here" and "Last 5 seconds".
- Confirm with "Done". "Undo" restores it at any time.
- The audio file is not cut. Lines outside the range are only hidden.
- The Transcript screen doesn't show the lines outside the range, and shows one line at the edge saying the start has been trimmed, with "Restore". The player plays from the start of the range and stops at its end.
- Exports, AI proofreading, the card workshop, the transcript sent to myWasha, and the number of people and length in the list are all counted within the range.

9.12 Deleting recordings, the Trash, and reprocessing
Deleting
- When you delete a recording, it is not removed at once but moved to the Trash. You can restore it for 7 days.
- The confirmation says that "(title)" from (date and time) (N min) will be moved to the Trash, and that for 7 days it can be restored with "Undo" in the list or from "Trash" in Settings.
- The main button is "Keep", and the delete side is "Move to Trash".
- Right after deleting, the list shows "Moved "(title)" to the Trash." and "Undo".
- The Trash is in Settings on iPhone and Android (Section 13.5), and is "Recently Deleted" in the left column on Mac and Windows. Recordings older than 7 days are cleared when the app starts.
Reprocessing
A recording that has finished processing can be processed again from the row menu. Both options show, before you press, how many things will be lost (names, reassignments, text corrections) so you can check.
| Item | Details |
|---|---|
| Start Over… | Deletes all speaker names, reassignments, handling marks and text corrections, and regroups the speakers |
| Transcribe Again from Audio… | Redoes the transcription from the audio (for example when it was transcribed in the wrong language). It starts after you choose the number of people and the language again. Bookmarks remain |
Recordings that couldn't be processed, or that you stopped partway, show "Process Again…". A recording that stopped because the language data was missing can be processed from where it stopped with "Process Again" after downloading.
9.13 When processing doesn't go well
The "couldn't process" screen
- Heading "Couldn't process"
- "What happened" (one sentence)
- Where the recording is (e.g. "The recording is still on this device.")
- Buttons under "What to do next": "Try again", "Open Settings", "Export logs", "Choose another file", "Back to list", and so on
Main "what happened" messages (in meaning):
| Situation | Display (meaning) |
|---|---|
| Format can't be read | Couldn't read this audio format. Supported formats are m4a, mp3, and wav. |
| No speech recognition permission (iPhone) | Transcribing this recording needs speech recognition permission. Nothing has been processed yet. (→ Turn on speech recognition for Washa in Settings, then try again) |
| No language data | The data for (language) hasn't been downloaded. You can download it in Settings. |
| Transcription language data can't be made usable (iPhone, Mac) | Apple's speech recognition data for (language) couldn't be made usable on this device. … |
| Voice models for telling speakers apart are incomplete | Washa's voice models used for telling speakers apart and more aren't all on this device yet. … |
| No processing data (Android) | This device doesn't have the data needed to separate speakers yet, so the recording can't be processed. |
| Language not supported | Transcription in this language isn't supported. |
| Didn't work properly partway | Transcription or telling speakers apart didn't work properly partway through. |
When no lines were found
- "No lines found" and "This can happen when the audio is very quiet or has no speech.", with the buttons "Reprocess" and "Delete".
- If this appears on Windows, check the microphone privacy setting (Section 4.2).
When processing stopped partway
- "Processing interrupted", and "Stopped during "(stage)". The recording is still on this device."
- Example reasons: it stopped because you switched to another app; the time allowed to continue in the background was used up; the device ran out of memory and the OS closed the app.
- "Resume" continues from the saved checkpoint.
- On iPhone, when it stopped because you switched to another app, it continues automatically when you bring the app back to the front.

10. Identifying speakers, changing the number of speakers, and the Speaker library
Version checked: app build 59.
When processing finishes, speakers have provisional names like "Speaker A", "Speaker B"… ("話者A"… on the Japanese screen). Once you give them names here, the voices stay in the Speaker library, and from the next recording on, Washa suggests "Is this this person?".
10.1 The speaker panel
The screen for naming speakers and correcting a line's speaker is the same panel, wherever you open it from:
- When you press a speaker tag on the Transcript screen
- When you select lines and press "Change speaker"
- When you merge lines
- When you open the speaker identification flow with "Check speakers"
- When you press the speaker of a paragraph on an AI-proofread card (Section 12.5)
It appears as a sheet from the bottom on iPhone and Android, as the inspector to the right of the Transcript screen on Mac, and as a panel inside the window on Windows.
Top of the panel
- The line being asked about (the person who said "…") appears in a speech bubble.
- "Now: (current state)" and [Undo] appear. [Undo] reverts the panel's answers one at a time, newest first.
- "Listen to check" is the biggest button. Press it and the button turns into a progress bar while it plays that speaker's lines in a row.
Answers
The answers are in the same order on every panel.
| Answer | Details |
|---|---|
| (name) | People decided in this recording, or candidates from the Speaker library |
| New person | Give a name and add to the Speaker library |
| Someone else | A live voice that isn't part of the conversation (shop staff, passers-by) |
| TV or video | Played voices, such as TV, video or announcements in a building |
| Noise | Sounds that aren't voices, or misheard text. Not included in exports |
| Later | Move on without deciding |
| Person in the Speaker library | Choose from the Speaker library with "Search by name" |
- People with a close voice are marked "Sounds close". When there are many candidates, find the rest with "N more — search with 🔍".
- A panel opened from a line puts "Just this line" first. To apply it to all of that speaker's lines, choose "All N lines of (name)". The count is shown.
- When you answer, a line such as "Set to TV or video (this line)" and [Undo] appears at the top of the screen.
- While saving, "Saving…" appears. If it couldn't save, it shows "Couldn't save. The transcript hasn't changed."
- If you choose a person already decided, all of that speaker's lines move to that person. When one person's voice was split in two, choosing that person merges them into one.
- People with the same name who are different people can be kept apart ("Create as a different person").



10.2 The speaker identification flow
"Check speakers" on the Transcript screen opens a flow that decides the recording's speakers one at a time. On Mac it doesn't open automatically.
- Start screen: "There seem to be N people in this recording." It shows who spoke how much by the size of their faces ("Bigger faces spoke more"). If the number is wrong, fix it with "Wrong number of speakers". It doesn't proceed until you press "Get started".
- Question: Answer "Is this speaker (name)?" with "Yes", "No" or "Not sure (decide later)". "Listen to check" lets you listen. For a speaker with no candidate in the library, it asks "Who is this speaker?".
- After "No": "Then who is it?" lets you compare with people in the Speaker library and choose ("Most similar", "Named in this recording", "Others in your speaker library").
- Check: After you give a name and answer "Yes", it plays another line from the same speaker and checks "Is this (name) too?". Answer "Yes", "No" or "Later". On "No", "Who is it?" moves just that line to another person. Parts you didn't answer show as "N min not checked yet".
- Name: For someone not yet in the Speaker library, enter a name at "What's this speaker's name?" and press "Use this name". With "Add to speaker library" on, Washa looks for this person in other recordings too. With it off, the name is for this recording only.
- Summary: The last page, "Speaker summary", lets you check the results together. "Read the transcript" returns to the Transcript screen.
- Even if you close it, it continues from where you left off. "Continue identifying speakers" on the Transcript screen takes you back.
- When you gave the number of people but not enough people have names, it looks for them: "Is this speaker number N?".
- Even when there is no Speaker library yet (first use), this flow lets you name the first person.




10.3 Lines that go into the Speaker library
- Only lines that a person has confirmed go into the Speaker library. Unconfirmed lines are not added.
- When you decide a name, Washa compares with that person's past lines. If the voice is very similar to someone else, the name is given but the lines are not added to that person's Speaker library. In that case a message says the speaker was set to (name), but because the voice resembles (another person), it was not added to (name)'s Speaker library. To add it anyway, press "Add".
10.4 Speaker faces
- When a person in the Speaker library has no photo, a face is chosen automatically from 30 blurred illustrations of people, based on that person's lines. The same person always gets the same illustration in every recording.
- Press the face on a speaker card, or "Change picture", to open "Choose a picture". You can pick another illustration with "Pictures", or switch to a photo with "Choose photo". On devices with a camera, "Take photo" takes one on the spot.
- Right after choosing a photo, "Adjust photo" lets you drag to set the position and pinch to set the size (up to 4×).
- Faces also appear on this panel, in the bookmark list, on backchannel lines, on cards in Home and on shared pages.

10.5 Changing the number of speakers
When the grouping differs from the actual number of people (one person split into two, two people merged into one, etc.), regroup with "Change number of speakers" on the Transcript screen.
- The sheet title is "Change number of speakers". It shows the current breakdown (e.g. "3 speakers · Someone else 2 lines · TV or video 1 line · Noise 4 lines").
- When opened from the Transcript screen, choose the number with − / + and press "Regroup into N speakers". When opened from the speaker panel, press "Make it N". When the number is the same as now, the button can't be pressed, and it shows "Change the number to regroup."
- Explanation: regrouping carries over the speaker names you gave and the speakers you corrected by hand as much as possible.
- While regrouping, it shows that it is regrouping into N people.
- The result line has [Undo].
| Result | Example display (meaning) |
|---|---|
| Grouped | There are N now |
| Carried over | Carried over N names and M marked lines |
| Raised the number but not found | Couldn't find speaker N. There are M now |
| Confirmed lines | Kept the lines you confirmed for N people |
- When you raise the number, Washa also re-examines speakers whose names are decided, to find the missing person. Lines you confirmed are not moved. When it can't find anyone, it doesn't pretend it did.
- When you said 2 people but there are 3 voice groups, it asks "Which 2?" (listen with ▶ and choose the 2 people in the conversation). Groups you don't choose remain as a "Distant voice", and you can later answer Someone else / TV or video / Noise for them.
- People with very similar voices may not separate even when you give the number, and get mixed into one.
- Regrouping can't run during other processing or recording ("Regrouping can't start until other processing or recording finishes.").
- Changing the number makes an AI-proofread card out of date ("You changed the number of speakers, so the card is out of date"; Section 12.4).

10.6 Someone else, TV or video, and Noise
| Answer | Body | Export | Number of speakers | AI proofreading |
|---|---|---|---|---|
| Someone else | Stays | Appears as a line without a name | Not counted | Only the text is corrected |
| TV or video | Stays | Appears as a line without a name | Not counted | Only the text is corrected |
| Noise | Stays in faint text | Not included | Not counted | Not included |
- You can switch to another answer at any time on the panel, and revert with [Undo].
- These marks are carried over even when speakers are regrouped.
- Even if one person's voice is split into "Speaker A" and "Speaker C", deciding both as the same person makes them count as one person in the number, color, name and export.
10.7 The Speaker library and speaker cards
The Speaker library remembers the voices of the people you have named. People registered in it appear as candidates from the next recording on. Open the whole library from "Speaker library (N)" in "Speakers". Deleted speaker cards stay in "Deleted speaker cards" for 7 days and can be restored with "Restore to speaker library".
Speaker cards
One person's card in the library. Open it from "Speakers", or from "Open speaker card" on the panel.
-
Name (changing it also changes the name in recordings confirmed for that person) and face
-
"Lines used for matching": the list of lines used to suggest candidates. If another person's voice got mixed in, remove it from matching with "Not this person" (it is only removed; no replacement line is added)
-
"Merge with another speaker card": when the same person got split into two cards, makes them one
-
"Save lines to speaker library": a per-card switch. Decides whether lines from recordings confirmed from now on are added to this person
-
"Delete this speaker card": the confirmation asks whether to delete "(name)" from the Speaker library, noting it can be restored for 7 days. Buttons "Delete" and "Cancel"
-
After launch, the current Speaker library is also reviewed once. If lines were removed from matching, a notice appears at the top of "Speakers". "Restore" puts them back.
-
The Terms of Use (Article 4) require that, when registering a voice in the Speaker library, you explain the purpose and get the consent of the person being registered.
-
Your speaker card can be made in the first-run guide, or with "Create your speaker card" in Settings (Section 5.3).

Speaker colors
A person confirmed in the Speaker library gets their own color, the same in every recording (only when colors clash within one recording does it move to a free color). Unnamed speakers get colors by their order in the recording (10 colors are used and repeat after that). Unnamed speakers have a "?".
11. Logging in and myWasha
Version checked: app build 59. Logging in is available on iPhone, Mac, Android and Windows.
11.1 Logging in with a Teleport account
- Entry point: "Log in with Teleport account" at the top of Settings. On Mac, the "Account" tab in Settings, or the account row at the foot of the left column.
- Description: "For server features like AI proofreading. Recording and transcription work without logging in."
- "You can create one with a Google account. We ask for consent the first time you log in."
- Without logging in, recording, transcription, speakers and export work as before. Logging in is needed only for server features such as AI proofreading, the card workshop and myWasha.


11.2 The membership panel and Membership settings
Once logged in, it shows your face, your name, "Teleport account · (level)" and a tag with what is left.
| Item | Details |
|---|---|
| Manage in myWasha ↗ | Plan and what is left, campaign codes, history, news, release notes |
| Membership settings | Finishing style · alert when running low · log in with another account · log out · leave |
| Copy to myWasha | Section 11.3 |
Membership settings
- The "On this device" section: the finishing style for AI proofreading, and "Notify me when I'm running low".
- "Log in with another account" lets you log in again.
- "Log out": "Logging out keeps the recordings, transcripts, and speaker library on this device."
- The "Leave" section:
- "Stop using Washa ↗": deletes your records on Washa's servers. Your Teleport account remains.
- "Delete Teleport account ↗": deleted together with the records of all Teleport products.
- Both open the request page (myWasha) inside the app.



11.3 Copy to myWasha
While you are logged in, transcripts, speaker names, AI-proofread versions and cards are copied to myWasha. Audio is not sent (audio and face photos only for recordings you made a share link for; Section 11.4).
- Before the first copy, a notice "Your recordings will be copied to myWasha" appears once, and nothing is copied until you press "OK".
- You can stop it at any time with "Copy to myWasha" on the membership panel.
- "Delete everything on the server" moves everything copied to the trash and turns "Copy to myWasha" off. It stays in the server trash for 30 days.
- Cards edited in myWasha (on the web) come back to the app ("Brought in N edits from the web").
11.4 Share links for cards
- You can make a read-only link from the "Share" tab of the share and export sheet (Section 13.1) or from "Share" on a card.
- Choose "No password" (anyone who knows the link can read it) or "With a password" (asks for a password when opened; 4–64 characters), then "Create link".
- Right after making the link, the audio and photos are sent in turn. The sheet shows "Sending audio and photos (n / N). They appear on the shared page when done."
- Press "Copy link" and "Copied" appears.
- "Stop sharing" makes the link unopenable. "Delete from server" moves the myWasha copy to the trash (sharing stops too; it is deleted after 30 days). Even after deleting, you can make a new link.
- The shared page lists the faces and names of the people in it, and pressing a paragraph plays from there. The media bar at the bottom of the screen lets you play, skip 15 seconds and change the speed. Photos and videos taken while recording also appear as a cover and illustrations.
- Only for shared recordings, the audio and face photos are placed in myWasha (audio up to 100 MB; removed by stopping sharing, the trash, or deleting everything).
- When not logged in, it shows "Log in to share".
12. AI proofreading and cards
Version checked: app build 59. AI proofreading requires logging in. It is available on all 4 OSes.
12.1 Proofreading with AI
Press "Proofread with AI" on the Transcript screen, and the transcript text is proofread on Washa's server and turned into an easy-to-read "card" ("Turns this into a card with a title and a lead, easy to read.").
- What is sent is the transcript text, speaker names and terms. The recording's audio is not sent.
- When you press it, you choose in "How to proofread".
| Way | Details |
|---|---|
| Standard | Tidies the text so it reads well. |
| Detailed | Looks up names and terms on the web before tidying, and adds time-stamped marks such as "〔inaudible 12:34〕" where it couldn't be heard ("Looking up names (1–2 min)") |
- AI proofreading never starts on its own. Only recordings where you turned on "Proofread with AI automatically" on the speaker-count sheet are proofread right after processing (Section 9.2).
- When not logged in, it shows "Log in to proofread with AI" (AI proofreading is available after logging in; transcription and speakers work without logging in).

12.2 AI proofreading progress
A thin bar appears in the same place as the big "Proofread with AI". It is the same shape as the transcription bar.
| Status | Display |
|---|---|
| Queued | Waiting in line |
| Proofreading | Proofreading with AI · 42% · About 1 min left [Stop] |
| Finishing | Finishing the title and contents |
| Offline | Waiting for the connection (continues from where it was once connected) |
| Done | Turns into the card on the spot |
- Processing runs on the server, so it continues even if you close the app or go offline.
- "Stop" stops it at any time ("AI proofreading was stopped").

12.3 When it couldn't be done
When it couldn't be done, "Couldn't proofread with AI" appears with the reason and the next step.
| Reason | Display |
|---|---|
| Can't connect | Couldn't connect. |
| Login expired | Your sign-in has expired. (→ "Sign in again") |
| Allowance used up | You've used up this month's allowance. (→ "Open myWasha") |
| Recording that can't be AI-proofread | This recording can't be proofread by AI (too long or an unsupported language). |
| Only some paragraphs failed | Some paragraphs couldn't be proofread. The finished ones are kept. (→ "Retry the failed paragraphs") |
| The source was edited during proofreading | The source changed during AI proofreading, so this result wasn't applied. (→ "Try again" requests it again from the current transcript) |
| The source couldn't be handed over | Couldn't hand the current source to AI proofreading. |
| The finishing style couldn't be read | Couldn't read the finishing style chosen in Membership settings › Finishing style. |
| No lines to proofread | There's nothing to proofread (every line is noise or empty). |
12.4 Reading and redoing cards
- A recording that has been proofread shows the card when you open it.
- The title becomes the one AI gave (a name you gave yourself takes priority). Under the title come a summary of about 300 characters, headings and a table of contents, and notes only when needed.
- The table of contents floats; press it to jump to that section.
- Speakers are shown by face circles and colors, and pressing a paragraph plays from there.
- At the top of the card, the proofread card → things made in the workshop → "+" are lined up. The date row of each item shows the names of the people in it. Press a name and the row of people's tags opens on the spot.
- You can also edit the summary, headings and paragraph text ("Edit summary", "Edit heading").
Redoing
- If you edit the source text after proofreading, "You edited the source, so the card is out of date" appears at the top of the Transcript screen; if you change the number of speakers, "You changed the number of speakers, so the card is out of date". "Redo" remakes the card.
- "…" → "Redo the card…" can be pressed at any time. If you have edited the card, it checks "Your edits to the card will be lost. Redo it?".
- For an out-of-date card, export defaults to the source version.


12.5 Correcting speakers on the card
- When you correct a speaker on a card paragraph ("Which speaker said this paragraph?"), the answer is also applied to the source (the original transcript), and the Speaker library learns from it.
- "Undo" reverts the card and the source in one step, and [Redo] is offered too.
- Card paragraphs, like the source, can be set to "TV", "Other sound" or "Noise".
- When you correct speaker names, reassign lines or change the number of speakers in the source, the speakers of the card paragraphs change the same way. AI proofreading isn't redone, so there is no wait and no credits are used. Only when you edit the body text does "The source has changed" appear.
12.6 Card workshop (test) (iPhone, Mac)
Open "Card workshop (test)" with "+" ("Create card") at the top of the card, and you can send the transcript to Washa's server and make various cards ("Sends this transcript to myWasha and makes a card from it."). Logging in is required ("Log in to use this").
| Kind | Details |
|---|---|
| Translation | Readable with speakers and times. You can choose the target language |
| Slides | Swipe left and right |
| Slideshow | Advances automatically with playback |
| Manga | Choose the format (educational manga / 4-panel / a few pages) and the art style. Educational manga comes with an explanation after the manga. It takes 1–2 minutes to make |
- While making, it shows "Making (n / N)".
- Pressing a time or a play button inside a card plays it in the app's player. It can go full screen and be viewed in landscape too.
- Cards you made line up at the top of the card and on the "Your cards" shelf in Home.
- From the top right of the pane showing a card, you can make a share link for that card.
- If a speaker card has a face photo, that photo is also sent to the server and shown as the speaker's mark on the card. Manga draws the characters based on that photo.
- The list of cards you made and AI-proofread versions shows the credits used.
- This is a test version. The kinds you can make and how they look may change without updating the app.
12.7 Knowledge
The feature that adds AI-proofread cards to Knowledge is being tried inside Teleport and is not yet offered to the public. The screen shows "Not available in this beta yet".
13. Sharing, export and settings
Version checked: app build 59.
13.1 The share and export sheet
Press the share icon on the Transcript screen, and a single sheet, "Share and export", opens. On Mac ⌘E and on Windows Ctrl+E also open it (in that case it opens on the "Export" tab).
Top of the sheet
- The recording's title and what to output ("Proofread card", "This recording (source)", "This recording (source, before AI proofreading)")
- "Copy": choose from 3 kinds, "With time and names", "Copy text only" and "Copy as Markdown". Pressing it shows "Copied".
- Tabs "Share" and "Export"
The Share tab (place it outside Washa to show or add it; it keeps up when you edit)
- "myWasha link": a read-only link (Section 11.4). After AI proofreading, the link shows the card too.
- When not logged in, it shows "Log in to share a myWasha link and use Knowledge" and "Export and Copy work without logging in".

13.2 The Export tab
Hands over the contents at that moment as a file. Later edits do not carry over.
What to export
- "Card" (proofread text with title, summary and headings) or "Source version" (the transcript as is). The default is the card; when the card is out of date, it is the source version.
Formats
| Format | Details |
|---|---|
| Word | A Word file (.docx) with the title and date/time |
| Markdown | A Markdown file with the title |
| Captions + audio | A caption file (.vtt) and an audio file. Usable as captions with speaker names in video editing software and video players |
| Audio only | The audio file |
| RO-Crate (a bundle with photos and videos) | Bundles the audio, transcript, AI-proofread versions and photos and videos taken, with a time-stamped catalogue (iPhone, Android) |
- To hand over only the text, use "Copy" at the top of the sheet.
- For a recording without audio, it shows "This recording has no audio file, so captions + audio can't be shared."
- Unnamed speakers are written like "Speaker A". Lines marked as noise and lines without text are not included. Bookmarks are not included.
- For a trimmed recording, only the range is exported.
- PDF export is not in the current version.
- A preview of the "Transcript to export" appears at the bottom of the sheet. When long, only the beginning is shown, with "Showing the beginning only. The full text is included when you share or save."
How it is handed over
| OS | How |
|---|---|
| iPhone | The iOS share sheet ("Save to Files" also makes it a file) |
| Mac | A save window (for captions + audio, choose a folder) |
| Android | The Android share sheet |
| Windows | A save window, etc. |

13.3 Help improve Washa, and sending to the development team
At the very bottom of the Export tab is the "Help improve Washa" section.
Export to This Device (Washa Document)
- Description: it bundles this recording's transcript, the record of speaker corrections, and the audio into 2 files (Washa Document).
- Before you press, it shows "Audio (length) · about (size) (2 files)". A progress bar appears while exporting. If there isn't enough free space, it doesn't export.
- Press "Export to This Device" and 2 files are made (a file
….washa.jsoncontaining the transcript, the record of corrections and more, and the audio file). On iPhone and Android you can also "Share 2 files"; on Mac and Windows, "Save to folder". - Contents: title, date recorded, processing results, record of speaker corrections, live text (if any), and the place recorded (if any). Voices and features from the Speaker library are not included.
- Nothing is sent automatically. It leaves the device only when you share it yourself.
Send to the development team
You can send answer files (Section 18.2), or recordings that weren't separated well, directly to the Teleport Inc. development team. Press "Send to the development team" to open the sending screen.
- "This is for sending answer files and recordings where something went wrong. For other bug reports, questions and ideas, please use Discord."
- "What are you sending?": "Answer file" (I corrected who spoke and where each turn breaks myself) or "Problem" (write in the note around what minute and what happened).
- "Choose when to delete it": "When the beta test ends" or "After 1 year".
- "Please confirm before sending (both are required)": "It contains no confidential conversation" and "I decided myself that it's OK to send".
- Send with "Send" and a receipt number appears. For a recording already sent, it shows that it has already been sent, with the receipt number.
Teleport Inc.'s promises (the gist of the on-screen text):
- Used only to improve transcription and speaker separation. Never used to train AI.
- We cannot tell who sent it. There is no sign-in, and we receive no name, contact or account.
- Speaker names, title, place and time are removed from the transcript on this device before sending. Voices, what was said and the audio file are sent as is.
- Deleted at the time you chose (when the beta test ends, or after 1 year).
- Stored encrypted on Cloudflare, Inc.'s systems, which may be outside Japan.
To delete it, press "Delete what I sent" in this recording's share and export sheet, and it is removed right away. If you can no longer delete it — for example you removed the app or changed devices — it is deleted automatically at the time you chose. We can't delete individual items on request.
- Since the audio itself is included, send it only after getting the consent of the people recorded.
- The receipt number also appears in the recording's info (Section 9.10).

13.4 How to open Settings
| OS | How to open | Form |
|---|---|---|
| iPhone | "Settings" in the bottom bar | All sections on one page |
| Mac | ⌘, | A separate window. Tabs "Account", "General" (display, background), "Language" (screen language, transcription language) and "Logs" (guide section, export logs, usage data, Trash, About) |
| Android | "Settings" in the bottom bar | All sections on one page |
| Windows | Menu "File → Settings…" (Ctrl+,) | A separate window |
13.5 Settings items
The iPhone order (from the top):
| Section | Item | Details / options | Default |
|---|---|---|---|
| Account | Teleport account | When not logged in, "Log in with Teleport account". When logged in, face, name and what is left, plus "Membership settings", "Manage in myWasha ↗" and "Stop using Washa ↗" (Chapter 11) | — |
| Display | Appearance | Match system / Light / Dark | Match system |
| Display | Transcript text size | Small / Default / Large / Extra large / Largest | Default |
| Language | Screen language | Match system / Japanese / English | Match system |
| Language | Transcription language | Installed languages, downloadable languages, main language (Section 7.4) | From the device language |
| Camera while recording | Take photos and videos while recording / Also save to the Photos app | Section 7.10 (iPhone, Android) | Take: on / Photos app: off |
| Background | Choose / change / remove a background image | Photo | None |
| Guide | Storage location (Mac, Windows), Show the introduction again, Create (or redo) your speaker card, Release notes | — | — |
| Logs | Export logs | Section 13.6 | — |
| Usage data | Send usage data | "We never send recordings or text, only counts of how you use the app." | On |
| Trash | Restore, Empty Trash (shown only when there is something in it) | Section 13.9 | — |
| Updates | Check now (Android, Windows) | Section 4.4 | — |
| About | Version, device, models and licenses, Terms of Use, Privacy Policy | — | — |
- Transcript text size is applied on top of the OS text size setting. Check it with the sample "Transcripts look like this." On Mac and Windows, the text size setting applies to the whole screen.
- With screen language "Match system", whichever of Japanese and English comes first in the device's language list is used. If neither is there, English.
- Background: shown on the recording list and in Settings; the Transcript screen stays plain. The image is kept only on this device.
- Storage location (Mac, Windows): "Show in Finder" ("Show in File Explorer" on Windows) and "Change location". Changing the location does not move the contents. It can't be changed while processing or recording. Washa reopens with the new storage location.
- Release notes: you can read what changed in each version ("What's new"). They are not shown automatically after an update. The last day of the beta also appears here.


13.6 Export logs (for bug reports)
- "Export logs" in Settings, or "Export logs" in a recording row's menu, writes the activity log to a text file.
- Description: "If something isn't working, send these with your bug report. They don't include audio, transcripts, or people's names."
- Contents: app version, OS, device model, free space, recording IDs, processing status, and so on. Audio, text, people's names and titles are not included.
- Nothing is sent automatically.
- Please report problems on the Washa Discord. Write the OS you use, the app version (the line under the logo, "0.2.0 · build N" or "1.0.0 · build N"; Section 13.8), and what you did and what happened, and attach this log if needed. For a problem with one recording, pasting "Copy Info" from "Recording Info" (Section 9.10) makes things certain. If you sent a recording with "Send to the development team", include the receipt number too (Section 13.3).
13.7 Models and licenses in use
Under "About" → "Models and licenses" in Settings, you can read the names, authors, licenses and full license texts of the AI models and components the app uses. On Mac it also opens from "License" in the app menu, and on Windows from the menu "Help → License".
- The information shown here is public.
- Anything beyond that, such as how the models are used or combined, is not public (Section 2.5).
13.8 How to check the version
- The version line appears under the logo: at the top of the list on iPhone and Android, and at the top left of the window on Mac.
| OS | Example |
|---|---|
| iPhone, Mac | 0.2.0 · build 59 |
| Android, Windows | 1.0.0 · build 59 |
- The first number is the app version, and "build 59" is the serial number shared by all OSes. The version number differs by OS, but the same build number means a version from the same point in time.
- "Version" under "About" in Settings shows a detailed line including symbols. When reporting a problem, attaching this line as is makes things certain (the "Version" in Settings can be selected and copied).
- If it shows "build dev", it may not be a version distributed for the beta. Please let us know on the Washa Discord.
- "Device" in Settings shows the device and OS version (e.g. "iPhone (iOS 26.5)", "Mac (macOS 26.x)", "Pixel 8a (Android 16)", "Windows 11").
- The top of the release notes also shows something like "v0.2.0 build 59 — Build 59".
13.9 Trash
- iPhone and Android: A "Trash" section appears in Settings only when there are recordings in the Trash. "Trash (N recordings; cleared after 7 days)", and for each row the title, "Moved on (date and time) · N min", and "Restore".
- Mac and Windows: Deleted recordings stay in "Recently Deleted" in the left column for 7 days. Rows show "Restore". When empty, "No recently deleted recordings".
- "Empty Trash": the confirmation says the N recordings in the Trash will be deleted from this device and can't be restored. The main button is "Keep", and the other is a red "Delete all".
- Recordings past 7 days are cleared automatically at launch. When free space runs short before recording starts, they are also cleared, oldest first (Section 7.1).

14. Data handling, Terms of Use, Privacy Policy
Version of the Terms of Use and Privacy Policy: 2026-10-21 edition (established October 6, 2026; revised October 21, 2026). The authoritative texts can be read on the in-app agreement panel and under "About" in Settings. This chapter summarizes the key points; if they differ, the in-app texts prevail.
The Terms of Use and Privacy Policy also describe a "web version" of Washa besides the app. The beta test covers the apps on the 4 OSes.
14.1 Where data is kept (summary)
| Data | Where it is kept | Does it leave the device? |
|---|---|---|
| Recordings (audio) | On the device | Only when the user shares, exports or sends it to the development team themselves, and when a share link is made |
| Transcripts (transcription, speakers, title, date, processing record) | On the device | When logged in and using Copy to myWasha, AI proofreading or the card workshop (Section 14.2), and when the user shares or exports it themselves |
| Recording location (place name and approximate position) | On the device (only if location is allowed) | When turning it into a place name, the approximate position is sent once to the OS service |
| Photos and videos taken | Inside that recording (on the device) | When put out through a share link or export |
| Speaker library (names, photos, voice features, audio of registered lines) | On the device | Voice features never leave. Face photos are sent when you use share links or the card workshop |
| Settings (appearance, text size, list order, record of agreement) | On the device | Never |
| Activity log (processing stages and failure reasons; no audio, text, names or titles are written) | On the device | Only when the user shares it with "Export logs" |
- If you are not logged in, the basic features are complete on the device.
- There is no feature that syncs recordings or the Speaker library directly between the apps on your devices. The Speaker library is separate for each device.
- iPhone and Mac do not sync through iCloud Drive. However, data may be included in iCloud Backup or Time Machine by the OS.
- On Android, data is excluded from backup and device transfer. It is not carried over when you change devices.
- On Windows, data is kept in the per-user local app data location (or the storage location chosen in Settings).
- Deleting the app deletes the data on the device. Please keep your own copies (Terms of Use, Article 5).
- For improvement, usage data (features used, counts, time, version, OS) is sent. Audio and anything made from it are not sent. You can stop it with "Send usage data" in Settings. There is no advertising mechanism.
14.2 When the app communicates outside the device
| Situation | What is sent | Recipient |
|---|---|---|
| Downloading the models used for processing | Only an ordinary download request (IP address, etc.) | iPhone, Mac: Hugging Face for speaker separation, Apple for speech recognition. Android, Windows: Hugging Face and GitHub |
| Downloading new versions (Android, Windows) | Only an ordinary download request | The distribution site |
| Converting to a place name (if location is allowed) | The approximate position, once when recording starts | iPhone, Mac: Apple. Android: the device's location service (Google Play services) |
| AI proofreading (used after logging in) | Transcript text, speaker names and terms. The recording's audio is not sent | Washa's server |
| "Detailed" AI proofreading | In addition to the above, names that appear in the recording and a summary of the topics, sent for searching | Washa's server and a search service |
| Copy to myWasha (while logged in) | Transcripts, speaker names, AI-proofread versions and cards. Audio is not sent | myWasha |
| Share links | The audio of the shared recording (up to 100 MB) and face photos | myWasha |
| Card workshop | The transcript and face photos from speaker cards | Washa's server |
| Send to the development team | All of the recording's audio, and the transcription with names, title, place and time removed | Teleport Inc. (stored on Cloudflare) |
| When the user sends something themselves | What they chose (exports, Washa Document, logs) | The recipient the user chose |
| Usage data (can be stopped in Settings) | Features used, counts, time, version, OS, and an anonymous per-device identifier | PostHog (USA) |
- Teleport receives recordings only when the user performs a sending action themselves, or uses server features after logging in (Terms of Use, Article 5).
14.3 Key points of the Terms of Use (2026-10-21 edition)
| Article | Key points |
|---|---|
| Article 1 Application | The terms apply to the whole relationship between the provider and the user. Individual notices within the app are also part of the terms |
| Article 2 Definitions | Defines recording data, processing results (transcription, speaker segments, names, voice features, glossaries, etc.), user data, and external services |
| Article 3 License | Grants a non-transferable, non-exclusive right to use it on your own device or account. Third-party models and components follow their respective licenses. Test versions (TestFlight, etc.) may change features or display without notice |
| Article 4 Consideration for people being recorded | Obtain the person's consent in accordance with the law before recording or processing. Do not infringe honor or privacy. When registering a voice in the Speaker library, explain the purpose and obtain the consent of the person being registered |
| Article 5 User data | Rights remain with the user. The basic features can be completed on the device without logging in. If you join Teleport's service, log in and use server features (storing and syncing recordings and transcripts, AI proofreading, translation, card creation, etc.), we receive and store the recording's audio, transcription, transcripts and so on on our servers for those features. With features that use external services, such as AI proofreading and translation, the necessary data is sent to external services. Users keep their own copies of the data. The user who sent something with "Send to the development team" is responsible for its contents |
| Article 6 Prohibited acts | Acts contrary to law or public order and morals, recording or processing without consent, analysis by decompiling, etc. (except as permitted by law), copying, modification, redistribution, sale, impersonation, etc. |
| Article 7 Intellectual property | Rights to the app belong to the provider or the legitimate rights holders |
| Article 8 Changes, suspension, termination | May be done without notice. When there is a significant impact, notice is given in advance as far as possible |
| Article 9 Disclaimer of warranties | Provided as is, with no guarantee of the accuracy of transcription, speaker segmentation or identifying people. Processing results are machine estimates, so users should check them before important uses |
| Article 10 Limitation of liability | No liability except in cases of the provider's intent or gross negligence. Where the Consumer Contract Act or similar does not allow the exemption, liability is limited to ordinary and direct damages, capped at the amount paid in the 12 months before the damage (10,000 yen if nothing was paid) |
| Article 11 Suspension of use | Use may be suspended without notice in case of violation |
| Article 12 Changes to the terms | May be changed in accordance with Article 548-4 of the Civil Code. Agreement is asked for again the first time you use the app after a change |
| Article 13 Language | There are Japanese and English versions; if they differ, the Japanese version prevails |
| Article 14 Governing law and court | Japanese law. The Tokyo District Court has exclusive agreed jurisdiction in the first instance |
| Article 15 Inquiries | Teleport Inc. (Bunkyo-ku, Tokyo) |
- There is no article about age.
14.4 Key points of the Privacy Policy (2026-10-21 edition)
| Item | Key points |
|---|---|
| 1. Basic approach | The basic features (recording, transcription, speaker segmentation, the voice library, etc.) are processed on the device, and user data is not sent unless you use server features. If you log in and use server features (storing and syncing recordings and transcripts, AI proofreading, translation, card creation, sharing and joint editing in the browser), we receive and store the recording's audio, transcription, transcripts and so on for those features. For improvement, we receive usage data (not including audio or its contents). With AI proofreading that uses web search, names that appear in the recording and a summary of the topics are sent to an external service for searching |
| 2. Information we receive | Account information, data the user sends (data when using server features, diagnostic logs, recordings sent with "Send to the development team", inquiries), and usage data (can be stopped in the app's settings). The app communicates to download models and new versions and to convert to place names, but no user data is sent |
| 3. Purposes of use | Providing features, managing logins and the scope of sharing, investigating problems and answering inquiries, improvement and developing new features, notices, and dealing with violations of the terms. Not used to train machine learning unless separately agreed |
| 4. External services and outsourcing | Storage or processing may be outsourced, or data needed for a chosen feature may be sent to external services (e.g. Google Cloud, Cloudflare, OpenAI, PostHog; may be outside Japan, such as the USA) |
| 5. Provision to third parties | Not provided without consent, except as required by law |
| 6. Retention period | Kept for the period needed for the purpose, and deleted when no longer needed |
| 7. Security measures | Measures such as encrypted communication and limiting who can access are taken |
| 8. Disclosure, correction, deletion | Data on the device can be deleted within the app. Data stored on our servers can be deleted from the app or the web screens. For disclosure, deletion, etc. of personal data we hold, contact the inquiry desk |
| 9. Changes | Important changes are announced and agreement is asked for again. The Japanese version prevails |
| 10. Inquiries | Teleport Inc. (Representative Director: Tomoyasu Hirano) |
14.5 Common questions about data
- "Is my recording sent anywhere?" → If you are not logged in, no. Processing happens on the device. If you log in and use AI proofreading or myWasha, transcripts and so on are received on the server (no audio is sent for AI proofreading or Copy to myWasha; making a share link places that recording's audio). Something reaches the development team only when you press "Send to the development team" yourself.
- "Where is my voiceprint?" → Only on the device, as the Speaker library. It is not synced with other devices either. When you delete a speaker card, the voice features and registered lines are also deleted after 7 days.
- "Can I record other people's voices?" → The Terms of Use require you to obtain the person's consent, in accordance with the law, before recording or processing. When registering a voice in the Speaker library, also explain the purpose and get consent. When sending to the development team, also think about whether it could cause trouble for the people recorded before you send it.
14.6 For hospitals, companies, and local governments considering Washa
Answers to questions we often receive from people considering Washa for a hospital or a company. Current as of November 12, 2026 (build 113, the build number shared by all OSes).
The key points
- If you do not sign in, recordings, transcripts, speaker names, and voiceprints never leave the device. All processing happens on the device.
- Voiceprints (voice features) never leave the device, even when you are signed in. They are not synced to other devices either.
- If you sign in and use server features such as AI proofreading, the transcript text and speaker names leave the device (to myWasha and OpenAI). The audio does not.
1. What leaves the device, by way of use
| Way of use | Audio | Transcript and speaker names | Voiceprints |
|---|---|---|---|
| Not signed in | Does not leave | Does not leave | Does not leave |
| Signed in, using AI proofreading | Does not leave | Leaves (to OpenAI, through myWasha) | Does not leave |
| Signed in, copying to myWasha | Does not leave | Leaves (stored in myWasha) | Does not leave |
| Creating a share link | Only that recording leaves (stored in myWasha) | Leaves | Does not leave |
| Pressing "Send to the development team" | Only that recording leaves | Leaves, with names, titles, places, and times removed | Does not leave |
Besides these, the app communicates outside the device in the following cases (see Section 14.2 for details).
- Usage data: features used, counts, time, version, and OS. It contains no audio and nothing made from it. It is sent by default, and each user can stop it per device with "Send usage data" in Settings.
- Downloading models and new versions: fetched on first use and similar times. No user data is sent.
- Converting a location to a place name: only if location access is allowed, the approximate location is sent once to the OS service when a recording starts.
- Account information: only when signed in, we receive your name, email address, and similar information.
2. Which companies process data, and in which countries
| Processing | Company | Country |
|---|---|---|
| Recording, transcription, speaker separation, voiceprints | (on the device) | — |
| AI proofreading (proofreading and the web search in "Detailed") | OpenAI | United States |
| Passing AI proofreading requests and answers; myWasha (storage, sync, sharing) | Cloudflare | Not fixed |
| Storing recordings sent with "Send to the development team" | Cloudflare | Not fixed |
| Usage data | PostHog | United States |
- We do not use user data to train machine learning models (Privacy Policy, Section 3).
3. What an administrator can control (current version)
| What you want | Current version |
|---|---|
| Keep users from using server features (such as AI proofreading) | They cannot be used without signing in. However, an administrator has no setting to turn them off for everyone |
| Stop sending usage data | Each user stops it per device. An administrator has no setting to turn it off for everyone |
| Sign in with an organization account (university or company) | Not available. Sign-in is only with a Teleport account (created with a Google account) |
| Let an administrator see who used the app and when | Not available |
We have received requests for a version for organizations (administrator settings, organization accounts, and usage records). No release date has been set.
4. Japan's medical information security guidelines ("3 ministries, 2 guidelines")
For the security of medical information in Japan, there are the Ministry of Health, Labour and Welfare's "Guidelines for the Security Management of Medical Information Systems" (version 7.0, June 2026) and the Ministry of Internal Affairs and Communications and Ministry of Economy, Trade and Industry's "Guidelines for Safety Management by Providers of Information Systems and Services Handling Medical Information" (version 2.0, March 2025). Our view for each way of use is summarized below. Whether Washa can be used is for each medical institution to decide.
| Way of use | Our view |
|---|---|
| Used only on the device, without signing in | Medical information never leaves the device, and we do not receive it. There is also no operation or maintenance contract between us and the medical institution, so we consider that we are not a provider covered by the MIC/METI guidelines (Section 2.1 of those guidelines). The devices and the app are managed by the medical institution as part of its medical information systems, in line with the MHLW guidelines (a device register, rules for personal devices, screen locks, and so on) |
| Using AI proofreading, myWasha, or share links | Medical information would be processed and stored outside the institution, and we would be covered by the guidelines. In the current version, do not use these features with recordings that contain patient information. We have not yet prepared a service specification conformity disclosure (MDS/SDS). myWasha's storage is not limited to Japan (the guidelines require that equipment storing medical information be within the reach of Japanese law) |
- "Medical information" in the guidelines means information that includes patient information. Recordings of meetings where patients are not discussed (such as department meetings or administrative meetings) do not fall under it. However, please treat attendees' names and similar details as personal information.
- Before recording, obtain the consent of the people being recorded (Terms of Use, Article 4).
15. Known issues and limitations
Each item in this chapter is written together with the version it was confirmed on. If your version is newer, it may have been fixed.
| # | OS | Issue | Workaround | Version confirmed |
|---|---|---|---|---|
| K-1 | Android, Windows | In short recordings or recordings of one person speaking, speakers may be split into more than there really are | Choose the correct number on the speaker-count sheet before processing. After processing, regroup to the correct number with "Change number of speakers" | Confirmed on the version just before build 14 (2026-10-05). Not rechecked on build 59 |
| K-2 | Android | Some processing data may be incomplete, showing "This device doesn't have the data needed to separate speakers yet, so the recording can't be processed." and processing can't run | Connect to Wi-Fi and wait a while. If it keeps appearing, report it on the Washa Discord | build 14 |
| K-3 | Windows | If Windows' microphone privacy setting is off, silence is recorded without an error | Turn on "Let desktop apps access your microphone" under "Settings → Privacy & security → Microphone" | build 14 |
| K-4 | Mac | Sound playing on the computer (system audio) can't be recorded. Microphone only | For online meetings, let the microphone pick up the sound from the speakers | build 14 (limitation) |
| K-7 | All OSes | Even when you ask for a number, it may not be possible to split into that number (people with very similar voices get mixed into one; when not found, "Couldn't find speaker N. There are M now") | Reassign speakers line by line on the speaker panel | build 14 (by design) |
| K-10 | iPhone | Operation on the iOS 27 beta has not been checked | Use iOS 26 | build 14 |
| K-11 | iPad | Same screens as iPhone. No screens made for iPad | — | build 14 (by design) |
| K-12 | Windows | Windows 10, ARM versions of Windows, USB microphones, and key operations with a physical keyboard have not been checked on real devices | Use Windows 11 (x64). Report any problems | build 14 |
| K-13 | Android, Windows | There was a report of English recordings coming out with each character separated by spaces | Choose "English" as the spoken language. If it doesn't help, report it | Reported before build 14. Not confirmed on build 14 |
| K-15 | Mac | Closing the lid stops recording | Keep the lid open while recording | build 14 (by design) |
| K-16 | Windows | Closing the "…" menu or the "Trim start and end" sheet with Esc closes the Transcript screen too | Close it with the sheet's close button, not Esc | build 59 |
| K-17 | Windows | The card's "…" doesn't show the Version section, "Redo the card…" or Photos taken | — | build 59 |
- K-5, K-6, K-8, K-9 and K-14 (listed as of build 14) have been removed because they were fixed or the behavior changed by build 59.
16. Differences by OS (table)
Version checked: app build 59.
| Feature | iPhone | Mac | Android | Windows |
|---|---|---|---|---|
| Recording | ○ | ○ (microphone only) | ○ | ○ |
| Live text during recording | ○ | ○ | ○ | ○ |
| Keeps recording with the screen off | ○ | (stops on sleep) | ○ | — |
| Recording/processing display on the lock screen | ○ (Live Activity) | — | ○ (notification; chip too on Android 16 and later) | — |
| Control Center / Quick Settings | ○ (also the Action button) | — | ○ (tile) | — |
| Home screen widgets (recording, to check) | ○ (also on the Lock Screen) | — | ○ | — |
| Shortcuts app | ○ | — | — | — |
| Camera while recording | ○ | — | ○ | — |
| Recording location | ○ | ○ | ○ | Manual entry only |
| Import from files | ○ | ○ | ○ | ○ |
| From Voice Memos (guide) | ○ | ○ | — | — |
| From other apps' Share | ○ | ○ | ○ | — |
| Drag and drop | — | ○ | — | ○ |
| Transcription languages | 10 | 10 | 5 | 5 |
| Screen layout | Bottom bar | 3 columns | Bottom bar | 3 columns |
| Speaker-count sheet (before processing) | ○ | ○ | ○ | ○ |
| Speaker panel, Speaker library | ○ | ○ (inspector) | ○ | ○ |
| Change number of speakers (1–10) | ○ | ○ | ○ | ○ |
| Editing lines (text, split, merge, delete, change speaker) | ○ | ○ | ○ | ○ |
| Find and replace | ○ | ○ | ○ | ○ |
| Trim start and end | ○ | ○ | ○ | Being checked |
| Recording info | ○ | ○ (⌘I) | ○ | ○ (Ctrl+I) |
| Bookmarks | ○ | ○ (⌘D) | ○ | ○ (Ctrl+D) |
| Export (Word, Markdown, captions + audio, audio only) | ○ | ○ | ○ | ○ |
| Export (RO-Crate) | ○ | — | ○ | — |
| Login, myWasha | ○ | ○ | ○ | ○ |
| AI proofreading | ○ | ○ | ○ | ○ |
| Card workshop (test) | ○ | ○ | Being checked | Being checked |
| Send to the development team | ○ | ○ | ○ | ○ |
| Menu bar, keyboard operations | — | ○ | — | ○ |
| Form of Settings | One page (bottom bar) | Separate window (4 tabs) | One page (bottom bar) | Separate window |
| Changing the storage location | — | ○ | — | ○ |
| How new versions arrive | TestFlight | TestFlight | Updates itself | Updates itself |
| Appearance, text size, screen language, background | ○ | ○ | ○ | ○ |
| Processing notifications | Notifications + Lock Screen + in-screen notices | Notifications + in-screen notices | Notifications | In-screen notices |
17. Frequently asked questions
Frequently asked questions and answers. The basis for each answer is in the relevant chapter.
Pricing and availability
- Q. Is it really free? Will it become paid after the beta ends? A. The basic features (recording, separating speakers, and using the text of who said what) remain free even after the beta ends. There is no monthly fee and no cap on time. Server features such as AI proofreading are used after logging in, and usage is counted in credits (Chapter 3).
- Q. Where can I see my AI proofreading plan and what is left? A. Open it from "Manage in myWasha ↗" on the membership panel (Section 11.2).
- Q. How many hours can I use it? A. There is no cap. Processing is done on the device, so there is no limit on the length of a recording or on monthly use (within what the device can handle).
- Q. How many people can it tell apart? A. Up to 10 per recording.
- Q. Can't I use it without logging in? A. Recording, transcription, speakers and export work without logging in. Logging in is needed only for server features such as AI proofreading, the card workshop and myWasha (Chapter 11).
Accuracy
- Q. The speakers are wrong. A. If the number of speakers is wrong, use "Change number of speakers"; if it's just one line, press the speaker tag and reassign "Just this line" to another person on the panel. Choosing the correct number on the speaker-count sheet before processing improves the result (Section 9.2, Chapter 10).
- Q. Only one person spoke, but it was split into two or more. A. This is a known issue that tends to happen on Android and Windows (K-1). Choose 1 with "Change number of speakers".
- Q. TV sound or shop staff's voices end up as speakers. A. Choose "TV or video" or "Someone else" on the speaker panel, and they are no longer counted as speakers. Mark unwanted sounds as "Noise" and they are left out of exports too (Section 10.6).
- Q. It gets technical terms and names wrong. A. You can fix them with "Edit" → "Edit text" or "…" → "Find and replace". If you log in, "Detailed" AI proofreading looks up names and terms before tidying the text (Section 12.1).
- Q. How accurate is it? A. The basic processing runs on the device without cloud AI, so there are mishearings and speaker mistakes. Still, we believe it is at a level that's fine for practical use. Accuracy figures are not public.
Languages
- Q. What about meetings mixing Japanese and English? A. Choose "Auto (Japanese with some English)" as the spoken language.
- Q. What about meetings mixing Chinese and English? A. Apart from the Japanese–English combination, one recording is treated as one language. Choose the language mainly spoken.
- Q. Can I use French? A. Yes on iPhone and Mac. Not on Android or Windows (Section 7.4).
- Q. I can't download a 6th language on iPhone. A. On iPhone and Mac, up to 5 languages can be used at a time. Choose a language to remove and press "Remove and Download" (Section 7.4).
Data and privacy
- Q. Are recordings sent to the cloud? A. Not if you are not logged in. Where things are sent, and what, when you log in and use AI proofreading or myWasha is as in Section 14.2.
- Q. Can I view a recording made on iPhone on my Mac? A. There is no feature that syncs recordings directly between the apps on your devices. If you are logged in, transcripts and cards are copied to myWasha, so you can view them on the web. To use the audio on the Mac, send it there with "Audio only" in export, for example, and import it. The Speaker library is separate for each device.
- Q. Is my data carried over when I change devices? A. Not on Android. On iPhone and Mac it may be included in the OS backup, but Washa itself has no transfer feature. Export and keep the transcripts you need.
- Q. I want to delete my voiceprint (the Speaker library). A. Delete it with "Delete this speaker card" on the speaker card. It can be restored for 7 days; after that, the voice features and registered lines are also deleted.
- Q. What is location used for? A. To get the place where recording started just once and attach the place name to the recording. You can record without allowing it. Windows does not get your location.
- Q. I want to delete a recording I sent to the development team. A. Press "Delete what I sent" in that recording's share and export sheet, and it is removed right away (Section 13.3).
Operation
- Q. Is it OK to turn the screen off while recording? A. iPhone and Android keep recording. Mac stops when you close the lid. On Windows, closing the window saves and ends the recording (Section 7.8).
- Q. Can I use another app while processing? A. On iPhone, processing continues if it could be handed to the OS. Even if it stops, it resumes when you open the app. Android continues in the background. On Mac it continues even if you close the sheet. AI proofreading runs on the server, so it continues even if you close the app (Section 9.3, Section 12.2).
- Q. I want to restore a deleted recording. A. Within 7 days, restore it with "Restore" — from "Trash" in Settings on iPhone and Android, or from "Recently Deleted" on Mac and Windows.
- Q. Can I export to Word or PDF? A. You can export to Word (.docx), Markdown, and captions + audio (.vtt and audio). PDF is not in the current version (Section 13.2).
- Q. I want to remove the chat at the start of a recording. A. "…" → "Trim start and end" lets you set the range to use. The audio file isn't cut, so you can always restore it (Section 9.11).
- Q. I want to record the other side of a Zoom call on Mac. A. Computer sound (system audio) can't be recorded. Let the microphone pick up the sound from the speakers (K-4).
- Q. I recorded on Windows but no text appears. A. Check Windows' microphone privacy setting (K-3).
- Q. The new-recording key doesn't work on Mac. A. In build 59 it is ⌘N (it used to be ⌘R).
- Q. How do I install new versions? A. On iPhone and Mac they arrive through TestFlight. Android and Windows update themselves. You can also check with "Check now" under "Updates" in Settings (Section 4.4).
Concept
- Q. What is Washa aiming for? A. Washa is the "ears" through which AI agents (Telepotch) on Teleport's AI platform perceive this world (Chapter 2).
- Q. Who is Washa-kun? A. The AI agent (a Telepotch) in Washa. The idea is that, in features to come, Washa-kun understands the content and shapes it into various forms.
- Q. I want to bring WashaKit into my company. A. We intend to offer WashaKit in many places. The terms are not written in this document.
- Q. Can it recognize animal or bird sounds? A. Not in the current version. The future plan is to make it recognize natural sounds other than human voices too (Section 2.3).
18. How the beta test works
This chapter describes the plan as of 2026-10-29. The schedule may slip.
18.1 First beta and second beta
| First beta | Second beta | |
|---|---|---|
| Start | Started with build 16 | After the first beta. When it starts, we will let you know by email and on Discord (Section 18.4) |
| Suited for | People used to beta testing who can give detailed reports | A wide range of people who want to try it normally |
| Number of people | Up to about 100 at most (10–20 is also fine) | No limit planned |
| What we ask | Make and send recordings that become "answer files" (Section 18.2), and detailed reports. Testing on several devices and OSes if possible | Use it normally and give us your opinions |
| Thanks | Your name in the credits (for the first beta, we are considering things like larger text) | Your name in the credits |
- Since we distribute on all 4 OSes (iPhone, Mac, Android, Windows), in the first beta it helps a lot to have people who can test on several devices if possible.
- The details of how the credits are shown (where they appear, how names are written, etc.) are not decided yet.
- The beta works until December 31, 2026 (Section 4.5).
18.2 What we ask of first-beta testers: make "answer files"
What we most want to ask of first-beta participants is to make and send recordings that become "answer files" for Washa.
- Record (or import), and let it transcribe and separate speakers.
- In the Washa app, correct the result into the right form.
- Assign the speakers correctly (give names, reassign lines to the right speaker, change the number of speakers)
- Make where each turn breaks correct (split or merge lines that are wrong)
- Don't tidy the words; keep them exactly as spoken. A few mistakes left in are fine.
- When you have finished correcting, send it from "Send to the development team" at the very bottom of the "Export" tab in the share and export sheet, choosing "Answer file" (Section 13.3).
- When sent, a receipt number appears.
- When you send answer files, the development team can use them to keep improving Washa.
- Please send only audio that may be used for improvement. Since the audio itself is included, limit it to recordings for which you have the consent of everyone recorded. Please don't send recordings that contain confidential conversations.
18.3 What we ask of second-beta testers
- Use it normally and tell us your impressions and opinions.
- Examples of what we'd like to hear: the OS and device you used, the situation (meeting, interview, etc.) and the length and number of people of the recording, whether the speaker grouping was right, words it often misheard, how long processing took, operations that were hard to understand, problems, and features you want.
18.4 How to join (Discord)
- Everyday bug reports, questions and ideas are welcome on the Washa Discord.
- In the Discord role-assignment channel, press a reaction stamp.
- If you join the first beta: the hammer stamp → the channel for first-beta participants becomes visible.
- If you wait for the second beta: the beginner (young leaf) stamp → when the second beta starts, you get a notice by email and on Discord.
19. Glossary
| Term | Japanese | Meaning |
|---|---|---|
| Washa | 話者 | This app. Provided by Teleport Inc. |
| WashaKit | 話者キット | The transcription, voiceprint analysis and speaker separation engine inside Washa. Supports 5 OSes |
| Washa Document | 話者ドキュメント | A document format that lays out in time order what happened at the scene of a recording. In the app, made with "Export to This Device" |
| Telepotch | テレポッチ | The AI agents on Teleport's AI platform |
| Washa-kun | Washa くん | The AI agent (a Telepotch) in Washa |
| Episode | エピソード | One recording and its record. Also the name of the list of recordings |
| Home | ホーム | The screen with shelves of recent, Continue, your cards, news and suggestions |
| Bottom bar | 下の帯 | "Home / Episodes / ● REC / Speakers / Settings" at the bottom of the screen on iPhone and Android |
| Media bar | メディアバー | The playback bar that floats above the bottom bar during playback |
| Transcript screen | 読む画面 | The screen for reading, listening to and correcting the transcript or card of one recording |
| Source (source version) | ソース(ソース版) | The original transcript before AI proofreading |
| Card | カード | A transcript tidied up by AI proofreading, or something made in the card workshop |
| AI proofreading | AI 校正 | Proofreading the transcript text on Washa's server and making it a card. Requires logging in |
| Card workshop | カード工房 | Where translation, slide, slideshow and manga cards are made (test) |
| Credits | クレジット | How the amount used by server features is counted |
| Teleport account | Teleport アカウント | The account used to log in |
| myWasha | myWasha | The place on the web where transcripts and cards are copied, and share links and plans are managed |
| Speaker tag | 話者の札 | The tag with the speaker's name and color at the head of a line |
| Speaker panel | 話者を決める札 | The screen, the same in form everywhere, for naming speakers and correcting a line's speaker |
| Speaker library | 話者台帳 | The library that remembers the voices of people you have named |
| Speaker card | 話者カード | One person's entry in the Speaker library |
| Your speaker card | あなたの話者カード | A speaker card made with your own voice. Made in the first-run guide or in Settings |
| Identify speakers | 話者を決める | The flow of confirming speakers one by one while looking at candidates |
| Speaker-count sheet | 人数の紙 | The sheet for choosing the number of people and language (and AI proofreading) before processing |
| Change number of speakers | 人数を直す | Specifying a number and regrouping the speakers |
| Someone else | その他の人 | A live voice that isn't part of the conversation. Not counted as a speaker |
| TV or video | テレビ・映像の音 | Played voices such as TV, video and announcements in a building. Not counted as a speaker |
| Noise | ノイズ | Sounds that aren't voices, or misheard text. Not counted and not exported |
| Distant voice | 離れた声 | A voice group not chosen when you gave the number of people. Can be answered later |
| Live transcript | 録音中の文字 | Provisional text, without speaker separation, shown while recording |
| Text only | 字だけ | A recording without speaker separation |
| Bookmark | ブックマーク(栞) | A mark added to a line |
| Trim start and end | 前後を切る | Leaving the start and end of a recording out of the transcript. The audio file isn't cut |
| Interruption | 途切れ | A point during recording where the microphone became unavailable |
| Interrupted recording | 途中で止まった録音 | A recording that couldn't be finished. You choose from the list whether to use it or delete it |
| Export logs | 記録を書き出す | Making an activity log for bug reports |
| Send to the development team | 開発チームに送る | Sending answer files or recordings with problems to the development team |
| Receipt number | 受付番号 | The number shown when you send to the development team |
| Trash / Recently Deleted | ごみ箱 / 最近削除 | Where deleted recordings stay for 7 days ("Recently Deleted" on Mac and Windows) |
| Version line | 版の 1 行 | The display under the logo, such as "0.2.0 · build 59" |


