Embedded Captions

This composites captions into talking-head footage using actual subject matting, so text can sit behind the speaker. It ships with 36 visual identities across three engines: standard rail-plus-climax (most explanatory video), pure cinematic embed (hero typography throughout), and themed constitutions like neon signs written stroke by stroke or glitch decode effects. The default anchor mode is a clean verbatim rail with occasional promoted peaks, which is correct for most content. Embedding every word is the common mistake. Runs locally end to end with its own transcription and matting, no API keys. The tooling is mature and opinionated: it gates you at decision time, refuses multi-subject clips, and the step-by-step authoring flow keeps you from the typical overdesign traps.

Install to Claude Code

npx -y skills add heygen-com/hyperframes --skill embedded-captions --agent claude-code