What problem does it solve? Adding styled captions to talking-head video normally requires manual editing, keyframing, and masking work in a video editor. This Skill automates the full pipeline locally: it transcribes the speech, mattes the subject, and composites captions either as a clean lower-third rail or as typography embedded behind the subject, without altering the original footage. ## Core Features & Use Cases - 36-identity catalog: Pick one visual identity (e.g. anchor, cream, ink, neon, ordnance, terminal) from CATALOG.md; the engine, compiler, and authoring file are derived automatically. - Three rendering engines: Standard rail + embedded climax, pure Cinematic column-flow embeds, and Theme mode for VFX-grade themed compositions with plate reactions. - Local end-to-end pipeline: Whisper transcription, human subject matting (u2net/PP-MattingV2), safe-zone probing, deterministic compilation, preview-frame visual QA, and gated rendering to final.mp4. - Use Case: Given a 30-second founder update clip, probe the footage, pick the keynote identity, author a small JSON of caption choices, preview composite frames, and render a captioned video with the climax word embedded behind the speaker. ## Quick Start Add captions to my talking-head video clip.mp4 using the anchor identity and render the final captioned video.
