A Claude Code skill for transcribing audio files with speaker diarisation and automatic speaker name identification.

• Parakeet + FluidAudio (default): Fast, local transcription with speaker identification. Runs entirely on Apple Silicon (~5 min for 1 hour of audio).
• AssemblyAI (optional): Cloud-based, for non-English audio or when explicitly requested.
• Speaker names: Automatically identifies speaker names from filename, YouTube metadata, and transcript content.

• See SKILL.md for complete documentation and usage instructions.
• See REFERENCE.md for detailed reference notes (including filler-word cleanup behaviour).

# Run install
npx skills add HartreeWorks/skill--transcribe-audio

# When asked "Which agents do you want to install to?", select "Claude Code"
# in addition to the default "Universal" list.

If you get "command not found", install Node then try again.

Created by Peter Hartree. For updates, follow AI Wow, my AI uplift newsletter.

Find more skills at skills.sh and HartreeWorks/skills.