Transcript Summarizer

Turning a recording into show notes takes hours, and AI shortcuts bring a worse problem: quotes nobody said, timestamps that point nowhere, and links that were never mentioned.

This skill gives your AI a procedure that stays faithful to the transcript. It reads SRT, WebVTT or plain text, with or without timestamps, and writes a summary, chapters with timestamps, key quotes with times and speakers, show notes, the links and resources mentioned (as spoken, never invented), and social clip ideas with start and end times on sentence boundaries, a hook and the opening line.

Every quote is checked word for word by the bundled Node.js script before you see it. The script reports quotes that aren't in the transcript (with the closest real passage), quotes at the wrong time or from the wrong speaker, and punctuation-only differences. The same script turns SRT, VTT and timestamped text into clean JSON segments, removes repeated auto-caption lines and merges tiny cues.

Examples include a podcast SRT, a webinar VTT with auto-captions, an interview text export, a quotes file with a planted fake quote and the report that catches it, and finished show notes.

The transcript is the only source, so transcription mistakes carry through. Without timestamps there are no chapters or clips. It doesn't cut audio or video.