Washa: a tool that turns recorded conversations into text, speaker by speaker
It tells who said what and when, and keeps it as timestamped text. The speaker separation runs on your own device.
Features
What Washa can do
Separates speakers and lays them out in colors
It separates speakers by their voices. Even short replies keep who said them.
Suggests speaker names from voiceprints
Register voiceprints, and it suggests names from the people who were there.
Fix wrong names by picking from suggestions
Just pick from the suggestions to fix them later.
Export to Word
Use the transcript as a document right away.
Name suggestions were correct 93.2% of the time: the share of correct names in a 10-minute conversation between 2 people with registered voiceprints (measured by the Washa team, September 2026). This is not the accuracy of speaker separation itself.
Use cases
Where it’s used: meetings, interviews, one-on-ones, classes
When confidential talk shouldn’t go to the cloud. When you want to re-check who said something. When you want to spend less time fixing names afterwards.
Why we make it
Why Teleport makes Washa: so agents can remember records that know who spoke
Washa isn’t just a tool for convenient meeting notes. It keeps what was said along with whose words they were — so that an agent hearing it can get to know each person and become a companion who works with them. That’s why we built it to run on your device, so you can keep as much as you like. Washa is the very first form of Teleport.
We’re looking for first beta testers (free)
Builds and announcements are shared in our Discord community.
NEW