deep-transcribe 0.1.3

High-quality transcription, formatting, and analysis of videos and podcasts

deep-transcribe

Take a video or audio URL (such as YouTube), download and cache it, and perform a "deep transcription" of it, including full transcription, identifying speakers, adding sections, timestamps, and annotations, and inserting frame captures.

By default this needs API keys for Deepgram and Anthropic (Claude).

This is built on kash and its kash-media kit of tools for handling videos.

Usage

See the env.template to set up DEEPGRAM_API_KEY and ANTHROPIC_API_KEY.

uv tool install --upgrade deep_transcribe

# Pick a YouTube video, and do a basic, formatted, or fully annotated transcription:
deep_transcribe transcribe https://www.youtube.com/watch?v=ihaB8AFOhZo
deep_transcribe transcribe_format https://www.youtube.com/watch?v=ihaB8AFOhZo
deep_transcribe transcribe_annotate https://www.youtube.com/watch?v=ihaB8AFOhZo

Results will be in the ./transcriptions directory.

To run as an MCP server:

# In stdio mode:
deep_transcribe mcp

# In SSE mode at 127.0.0.1:4440:
deep_transcribe mcp --sse

Or for Claude Desktop, a config like this should work (adjusted to use your appropriate home folder):

{
  "mcpServers": {
    "deep_transcribe": {
      "command": "/Users/levy/.local/bin/deep-transcribe",
      "args": ["mcp"]
    }
  },
}

To debug MCP logs:

deep_transcribe mcp --logs

Project Docs

For how to install uv and Python, see installation.md.

For development workflows, see development.md.

For instructions on publishing to PyPI, see publishing.md.

---

This project was built from simple-modern-uv.