About Sayfield
A speaking-practice app, and a set of speech models in the making.
What it is
You answer a question out loud and get your recording back written down: the words, the ums, the pauses where they fall.
Sayfield is also building its own speech models, for the app and for people who build speech tools or study talk.
Where it stands
Private beta, invite-only. The app runs on open models with Sayfield's own rules on top. Sayfield's own models are in development; none has been released.
Who makes it
Sayfield is made by Abdurrahman Foudhaily.
The rules it keeps
- It describes the speech, never the speaker.
- Recordings from the app train its models only when the speaker says yes, as a separate choice.
- Its models will be tested against marks checked by people, never against another machine's raw output.
Contact
Found a security problem? Write to the same address.