Turn an audio or video recording into a timestamped transcript and structured meeting notes: summary, decisions, action items, follow-ups, and open questions.
- Save review time: turn long recordings into readable notes and searchable text.
- Keep the evidence: every transcript segment includes timestamps.
- Choose your AI provider: OpenAI, Anthropic, OpenRouter, DeepSeek, or a compatible custom endpoint.
- Export the result: download meeting notes, transcripts, or combined data as Markdown, TXT, or JSON; SRT is available through the API.
Useful for meetings, interviews, lectures, research recordings, and internal video updates.
- Upload an audio or video file up to 100 MB.
- Local
faster-whispercreates the timestamped transcript. - The selected provider turns the transcript into structured notes.
- Review the result, reopen it from history, or download an export.
Source media is removed after processing. The transcript and validated meeting notes stay in the local SQLite history until the job is deleted.
Provider keys are backend-only. They are loaded from
.env, never entered in the browser, and never stored in jobs or exports. Live analysis sends the transcript text to the selected external provider.
git clone https://github.com/snikmas/transcriber_app.git
cd transcriber_app
cp .env.example .env
# Add one provider key to .env, for example DEEPSEEK_API_KEY.
docker compose up --buildOpen http://localhost:8501. The API is available at http://localhost:8000/docs.
The first real transcription downloads and caches the configured Whisper model.
Run with Python instead
git clone https://github.com/snikmas/transcriber_app.git
cd transcriber_app
cp .env.example .env
uv sync --extra local
uv run uvicorn main:appIn a second terminal:
API_URL=http://127.0.0.1:8000 uv run streamlit run ui/app.py- Local speech recognition with
faster-whisper - Audio and video uploads with automatic video-to-audio conversion
- DeepSeek, OpenAI, Anthropic, OpenRouter, and custom provider support
- Structured summaries, decisions, action items, follow-ups, and open questions
- Timestamped transcript and durable local job history
- Markdown, TXT, and JSON downloads, plus SRT export through the API
- FastAPI, Streamlit, SQLite, Docker Compose, and 136 offline tests
Streamlit UI
│ upload + provider selection
▼
FastAPI API ─── SQLite job state
│
└── in-process worker
├── local faster-whisper transcription
└── configured external analysis provider
The project is designed for local or small-team use. See
docs/ARCHITECTURE.md for the data flow and
docs/PRIVACY.md for the trust boundary.
uv sync --extra local --group dev
uv run pytest -q
uv run ruff check .
uv run ruff format --check .
docker compose config --quietThe automated checks are offline: they do not call paid providers or use personal recordings.
Developer details: DEVELOPER_SUMMARY.md ·
RELEASE_NOTES.md
The release does not include authentication, speaker diarization, real-time capture, transcript editing, Zoom/Meet integrations, CRM sync, or distributed workers. It is a local/small-team application, not a hosted multi-tenant SaaS.
Speech recognition and AI-generated notes can contain mistakes. Review results before using them for decisions, commitments, or official records.
MIT. See LICENSE.