transcribe-meeting

Transcribe meeting recordings into speaker-attributed transcripts and structured briefs with action items.

1.6k|168|Updated Dec 16, 2024
One-click install
npx skills add https://github.com/amd/gaia --skill transcribe-meeting-amd
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: transcribe-meeting
Source: https://github.com/amd/gaia/tree/main/hub/skills/transcribe-meeting
Command: npx skills add https://github.com/amd/gaia --skill transcribe-meeting-amd

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Turning a long meeting recording into usable notes normally means hours of manual listening and typing, and naive summarization of a raw transcript loses speaker attribution and silently truncates content. This Skill converts audio or video recordings into a speaker-labelled transcript and a complete brief covering decisions and action items. ## Core Features & Use Cases - Three-stage pipeline: transcribe media to a raw transcript file, refine it into a speaker-labelled indexed markdown transcript, then summarize the full document into a detailed brief. - Speaker diarization with honest attribution: voices are separated acoustically and named from conversational cues, with unnamed voices kept as Speaker 1, Speaker 2 rather than guessed. - Indexed follow-up Q&A: the refined transcript is indexed so users can ask detailed questions about the meeting later, while only durable facts are stored in memory. - Use Case: Point the Skill at a 46-minute staff-meeting.mp4 and receive the raw transcript path, a speaker-attributed refined transcript, and a brief listing decisions, owners, and unresolved items. ## Quick Start Transcribe and summarize the meeting recording at C:\recordings\staff-meeting.mp4, including who said what and the action items.

Frequently Asked Questions about transcribe-meeting

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I transcribe a meeting recording and get a summary?

Provide the path to an audio or video file and the Skill runs three steps: transcribe_media produces a raw transcript file, refine_transcript creates a speaker-labelled indexed transcript, and summarize_document generates the brief with decisions and action items.

What audio and video formats can be transcribed?

The Skill accepts mp4, mkv, mov, m4a, mp3, and wav recordings, as well as Teams or Zoom transcript exports. Pass the file path directly; an optional language parameter skips auto-detection.

How long does transcribing a meeting take?

A measured 46-minute recording took about 6 minutes to transcribe, 4 minutes to separate voices, and 1-2 minutes to name speakers and summarize, roughly 11-12 minutes total or about 4x realtime.

Does the transcription identify who said what?

Voices are separated from the audio by acoustic diarization, and names are inferred from self-introductions and conversational cues. Voices that cannot be named stay labelled Speaker 1, Speaker 2, and are never guessed.

What happens if speaker diarization fails?

When voices cannot be separated, transcribe_media returns a speaker_identification field explaining why. The Skill reports this reason and still produces a brief, but without per-speaker ownership of action items.

Can I ask follow-up questions about a past meeting?

Yes. The refined transcript is indexed during refinement, so later questions about specific details are answered from the index via query_documents. Only durable facts like decisions and personal commitments are stored in memory.