meetnotes turns raw meeting recordings into concise summaries entirely on your own machine. It pairs local transcription with LM Studio for summarization, so sensitive meeting content never touches a third-party API. Multi-language prompt support and tunable context/output sizes make it easy to adapt to different meeting lengths and languages.
Use the helper script to run meetnotes inside the local virtual environment:
./meetnotes meeting.webmThe script creates venv if needed, installs requirements.txt, and forwards all arguments to meetnotes.py.
Set LM Studio context length for large meetings:
./meetnotes meeting.webm --context-size 32768When --context-size is set, meetnotes uses LM Studio's native /api/v1/chat endpoint because the OpenAI-compatible chat endpoint does not support per-request context length.
Set the maximum summary length separately:
./meetnotes meeting.webm --context-size 32768 --max-output-tokens 4096--context-size controls how much input and output the model can fit in memory. --max-output-tokens controls how long the generated summary is allowed to be.
By default, meetnotes uses prompts/default_en.txt.
Use another bundled prompt language:
./meetnotes meeting.webm --prompt-language ruBundled default languages: en, ru, de, fr, es.
Use a different defaults directory:
./meetnotes meeting.webm --prompt-dir ./my-prompts --prompt-language deUse a fully custom prompt file:
./meetnotes meeting.webm --prompt ./custom_prompt.txtPrompt files should contain {transcript} where the transcript should be inserted.