Roundup
Open-source speech-to-text on Android
As of October 2026, three open-source Android apps cover different dictation jobs. Sayboard and WhisperType are on-device keyboards. OSSFlow is a GPL-3.0 floating bubble that sends audio to an API you choose. Source you can read is not the same thing as an open-source licence.
Open-source apps checked for this page
OSSFlow
GPL-3.0 only, for this version. Android 14 or newer. A floating bubble pastes into the focused field. Transcription is Grok by default or any OpenAI-compatible endpoint over HTTPS. There is no on-device model, no account, and no OSSFlow server. Keys stay on the phone.
Sayboard
F-Droid lists Sayboard as an open-source on-device voice keyboard using Vosk. Suggested version v4.2.1 was added on 6 September 2024 and needs Android 6.0 or newer. You download a Vosk model. The listing says internet access is for that download and that your information is not sent.
WhisperType Keyboard
F-Droid lists WhisperType as MIT-licensed. It is a full QWERTY keyboard plus offline sherpa-onnx Whisper, with tiny.en, base.en as the default, and small.en int8. Dictation warns at 25 seconds and stops at 30. Version 1.5.3 was added on 21 September 2026 and needs Android 8.0 or newer. No API key. Models come from Hugging Face when you download them.
Cloud bubble or on-device keyboard
OSSFlow fits when you want the keyboard you already use, plus a bubble, plus a hosted model. You pay that host. The app itself is free, bring your own key. Audio leaves the phone. Read the privacy page before you rely on it.
Sayboard and WhisperType fit when the recording must stay on the device. They replace or become the keyboard. WhisperType's F-Droid page warns that the model download uses a non-free network service. After the model is on the phone, dictation is local. Sayboard's listing describes Vosk, not Whisper.
Source-available, and not on this list
FUTO Voice Input publishes its source, and the README says development has largely moved to FUTO Keyboard while Voice Input stays available. The licence is the FUTO Source First License 1.0. That text limits use and distribution to non-commercial purposes. It is not GPL, MIT, or another open-source licence, so this open-source list does not include it.
Other F-Droid speech packages exist. If their listing was not opened for this page, they are not here. Absence is not a verdict.
Questions
- Is OSSFlow fully on-device?
- No. The code is open source. The speech model is the API you configure. If you need offline open-source dictation, use Sayboard or WhisperType.
- Does Sayboard use Whisper?
- Not according to its F-Droid listing, which describes a Vosk voice keyboard and a Vosk model you download. Do not describe Sayboard as a Whisper app.
- Is FUTO Voice Input open source?
- No. The repository publishes source under the FUTO Source First License 1.0, which does not allow commercial use. Open source here means a licence that does not add that limit, such as GPL-3.0 or MIT.
- Which of these pastes into any app without becoming the keyboard?
- OSSFlow. Sayboard and WhisperType are input methods. You enable them as the keyboard. OSSFlow leaves the current keyboard in place.