Analysis updated 2026-05-18
Transcribe an audio file already saved on your phone into text without an internet connection.
Let another Android app, such as Forkgram, use Scrib in the background to turn voice messages into text on-device.
Download different sized Whisper models and switch between them to balance transcription speed and accuracy.
Build a compatible transcriber app using the open Open Transcribe contract instead of depending directly on Scrib.
| 23rd/scrib | composablehorizons/material-3-compose-unstyled | lyosu/hortay-android | |
|---|---|---|---|
| Stars | 14 | 14 | 14 |
| Language | Kotlin | Kotlin | Kotlin |
| Setup difficulty | moderate | moderate | hard |
| Complexity | 3/5 | 3/5 | 4/5 |
| Audience | general | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Building from source needs Android SDK 35, NDK 27, and CMake, plus cloning with git submodules for whisper.cpp.
Scrib is an Android app that turns spoken audio into text entirely on the phone, without sending anything to a server. It uses the open Whisper speech models to do the transcription, and the only time the app connects to the internet is to download a model the user chooses. Once a model is on the device, all transcription happens offline. There are two main ways to use Scrib. A person can pick any audio file already on their phone and have Scrib transcribe it directly inside the app. Scrib can also run quietly in the background as a shared transcription engine for other apps: an app such as Forkgram can send it a voice message and get text back, all still processed on the device, as long as that app lets the user pick Scrib as its offline transcriber. Users choose which Whisper model to download from the main screen, or let Scrib pick a suitable one based on their language. Bigger models tend to be more accurate but slower and larger to download, and a person can keep several installed and switch between them. Scrib also supports adding other whisper.cpp compatible GGML models by direct link or by importing a file already saved on the device. Models themselves are not included with the app and must be downloaded separately. For developers, Scrib exposes an open contract called Open Transcribe, a small Android service with no SDK and no network requirement, so any app can ask for offline transcription without being locked to Scrib specifically. Other transcriber apps that support the same contract could be used in its place. Building Scrib from source requires cloning with submodules, since its transcription engine is compiled from a bundled copy of whisper.cpp using CMake and the Android NDK. Because everything runs on the device, no voice recordings ever leave the phone, and network access is limited strictly to downloading a chosen speech model over HTTPS. Scrib is free software released under the GNU General Public License version 3.0 or later, and it bundles the separately licensed whisper.cpp project to run the underlying Whisper models.
Scrib is an Android app that transcribes speech to text entirely on the phone using offline Whisper models, so audio never leaves the device.
Mainly Kotlin. The stack also includes Kotlin, Android, Whisper.
GPL v3.0 or later: you can use and modify the code freely, but any modified version you distribute must also be shared under the same license.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Verify against the repo before relying on details.