fegger
404f3db200
- android/: Gradle/Kotlin project (AGP 9.4, Compose, NDK 27.1), monorepo subdir as planned; whisper.cpp v1.9.3 vendored via pinned fetch script - core/whisper: JNI wrapper (beam size, threads, language) + WhisperEngine Kotlin API mirroring the desktop engine contract - app (M0 scope): in-app GGML model download from Hugging Face, engine load, WAV picker with resampling, on-device transcription, share sheet - build validated: assembleDebug OK, libwhisper_jni.so + ggml packaged
MeetRec for Android
Native Android app with the same functionality as the desktop meetrec: record meetings and transcribe them on-device with Whisper. Nothing leaves the phone — no cloud, no telemetry.
Status: M0 (engine proof) — the app can download a GGML model, load it via JNI (whisper.cpp, CPU/NEON) and transcribe a picked WAV file. Recording, live transcript, and output files arrive in later milestones (see the milestone plan in the project docs).
Requirements
- Android Studio (or: SDK Platform 36, Build Tools 36, NDK 27.1, CMake 3.22.1)
- JDK 17+
- Device running Android 10+ (developed against a Fairphone 6 / Snapdragon 7s Gen 3, Android 15+)
Build
cd android
./tools/fetch-whisper.sh # vendors whisper.cpp v1.9.3 into third_party/
./gradlew :app:assembleDebug # or open the android/ folder in Android Studio
adb install -r app/build/outputs/apk/debug/app-debug.apk
third_party/whisper.cpp is gitignored — the fetch script pins the exact
release tag so the JNI layer never breaks on upstream churn.
Test on device (M0)
- Launch MeetRec, pick a model (
tinyis fine for a first test) and tap Download (model comes from Hugging Face; tiny is ~75 MB). - Tap Load engine.
- Tap Pick WAV file and choose a 16 kHz mono WAV for best results (other PCM WAVs are resampled automatically).
- Tap Transcribe — segments with timestamps appear; share via the share sheet.
Expect roughly realtime transcription with tiny/base on the
Fairphone 6's CPU; small is noticeably slower — use it for final passes
only (the live/final split comes with the recorder milestones).
Performance notes
- The engine runs on CPU via NEON, using up to 4 threads.
- GPU/NPU acceleration (e.g. Snapdragon NPU via the QNN backend) is a stretch goal, not wired up yet.
Module layout
app/ Compose UI (model download, WAV picker, transcript view)
core/whisper/ whisper.cpp JNI wrapper: CMake build + LibWhisper.kt +
WhisperEngine.kt (the on-device transcription API)
tools/ fetch-whisper.sh — vendor the pinned whisper.cpp release
Models live in the app's private storage (filesDir/models), shared files
with the desktop app's ~/.cache/meetrec/whisper-cpp naming
(ggml-tiny.bin … ggml-large-v3.bin).