skill

Describe a time range, scene, or output goal in Claude Code to cut footage, remove silence, turn long videos into Shorts, or generate and style subtitles. Changes are previewed through a dry run and backed by project backups and rollback guidance. The skill depends on the unofficial capcut-cli and CapCut or JianYing Desktop, so CapCut Web is unsupported and application schema changes may break compatibility.

talking-head-video turns a raw selfie or talking-head recording plus its script into a 1080×1920 vertical explainer. It transcribes word timings, corrects subtitle text against the script, replaces explanation beats with Remotion knowledge graphics, and keeps a circular speaker avatar in the top-left. An ffmpeg compositor burns the subtitles and overlays the graphics without altering the original audio. The workflow needs faster-whisper, Node with Remotion, ffmpeg, and a usable CJK font.

motion-mosaic-editing teaches an engine-agnostic method for adding motion graphics, kinetic captions, overlays, and infographics to existing footage. It decomposes an edit into reusable Tiles, keeps an element-addressable JSON scene specification, and moves from intake and analysis through tile design, captioning, reframing, scoring, export, and reuse. The skill explains how an LLM can decide while Remotion, SVG and Chromium, Canvas-WebGL, or FFmpeg perform the rendering. It ships methodology, not a renderer or proprietary Mosaic code.

Provide an approved script, talking-head or voiceover media, and gathered B-roll, and the skill guides Claude through a structured vertical-video edit. It applies A-roll/B-roll cutting, eased camera movement, themed motion graphics, sound design, and a brand style before HyperFrames renders the MP4. A bundled check verifies dimensions, duration, H.264/YUV420p output, and audio; Claude Code, HyperFrames, Node.js, and ffmpeg are required.