scriptor

What problem does it solve? Converting long audio or video recordings into accurate, structured meeting transcripts is tedious and error-prone, especially with domain-specific terminology and multiple speakers. This Skill automates the full pipeline from media file to final protocol while keeping the user in control at key review points. ## Core Features & Use Cases - Multi-model transcription: Supports Gemini (gemini-2.5-pro, gemini-2.5-flash) and OpenAI (gpt-4o-audio-preview, gpt-4o-transcribe) models with custom glossaries to improve recognition of domain terms. - Chunked processing with caching: Splits long recordings into 20-minute overlapping chunks via ffmpeg, transcribes each, aligns speaker labels across chunks, and resumes from where it stopped on re-runs. - Interactive post-processing: Thematic segmentation, terminology audit with user-approved corrections, participant identification, and final assembly into a structured transcript.md. - Use Case: You have a 90-minute meeting recording in Russian with technical jargon. The Skill converts it to mp3, transcribes it chunk by chunk with your glossary, lets you fix misrecognized terms, and produces a segmented protocol with a participants list. ## Quick Start Ask the agent to transcribe your recording by saying: transcribe the file meeting-2024-06-01.mp4 using my project glossary with gemini-2.5-pro.