- Category
- Video & audio
- Status
- In development
- Platforms
- iOS, Android
- Stack
- React Native, Expo, whisper.cpp
Speech-to-text using a quantised Whisper model executing locally. Recordings and transcripts never leave the device, which makes it usable for the conversations that most transcription services are disqualified from touching.
Transcription services are convenient precisely because they take your audio. ScribeZero does the opposite: a quantised Whisper model runs on the device, so the recording and the transcript stay there.
Where it actually stands
The engine is verified — a spoken fixture transcribes with complete word recall, checked against a reference transcript offline. The binding between the app and that engine is not yet fully exercised end to end, because the import path is gated and a simulator cannot be fed real microphone input. Until that is driven on a device, this one stays listed as in development. We would rather say so than ship a claim we have not tested.
- Local Whisper
- A quantised model runs on device; there is no API key and no upload step.
- Verified recall
- A spoken fixture is transcribed and diffed against a reference — the engine is measured, not assumed.
- Text and SRT export
- Output as plain text or as subtitles with timings.
- No account
- Nothing to sign in to, nothing stored off the device.
Captures of ScribeZero are pending. We would rather show you nothing than a drawing of a screen that was never photographed.