On-device audio transcription toolkit - a Python engine plus a small Android demo app. Trim, tag, transcribe and export your recordings. Everything runs locally: audio never leaves the machine.
I record a lot of voice memos and wanted one tool that could trim them, tag them, and turn them into searchable text without uploading anything to a server. WhisperDeck is that tool: a deck of recordings, transcribed on device with a local whisper.cpp build.
| Local-first | one-time model download is the only network call, ever |
| Trim + tag | wav trimming, VAD-based speech windows, sidecar tags |
| Transcribe | whisper.cpp wrapper with segment timestamps |
| Model zoo | tiny → large-v3, picked per clip length |
| Export | SRT, WebVTT and markdown tables |
| Android demo | foreground recorder service + deck browser |
whisperdeck/ the Python engine (CLI + library)
app/ minimal Android demo that records and lists a deck
models/ whisper model helpers and download script
docs/ guides (getting started, model zoo, shortcuts, exports)
examples/ end-to-end recipes (podcast notes, meeting minutes)
$ pip install -e .
$ wd deck list
$ wd transcribe memo.wav --model base.en
$ wd export deck.srt --format srtThe Android demo lives in app/ - open it in Android Studio, run it on a
device, and recordings land straight in a deck.
- Python 3.9+
- a
whisper.cppbinary on PATH (see docs/model-zoo.md) - Android Studio (only for the demo app)
PRs welcome - see CONTRIBUTING.md. Keep whisperdeck/
stdlib-only; run make test before pushing.
MIT - see LICENSE.