What problem does it solve? Recorded screen-cast walkthroughs contain false starts, repeated takes, and dead silence that require tedious manual editing. This Skill automates the transcribe-and-trim workflow so a raw recording becomes a tight, edited video without manual timeline work. ## Core Features & Use Cases - Transcription-driven cuts: Extracts audio with ffmpeg, transcribes with OpenAI Whisper at word-level timestamps, and proposes cuts based on narration cues like "scratch that" or repeated takes. - Collaborative review: Presents proposed cuts plus the resulting continuous transcript for user sign-off before rendering, with explicit flagging of judgment calls. - Visual cut editor: Generates a browser-based timeline editor served by a local Python server where segment boundaries can be dragged, previewed, and saved back to disk for re-rendering. - Use Case: Drop a 10-minute walkthrough recording into source/, ask for it to be cleaned up, review the proposed cuts, fine-tune boundaries in the browser editor, and receive a loudness-normalized final MP4 in output/. ## Quick Start Drop a video file into the source/ folder and ask Claude to clean up the video in source/.
