Routes your lipsync job across four RunComfy endpoints (Sync Labs v2/Pro, ByteDance OmniHuman, Kling, Creatify) based on what you're actually trying to do: dub an existing video, animate a portrait still, or generate speech from a script. The skill picks Sync Labs for mouth-swap on real footage, OmniHuman for talking-head avatars from a single photo, and Kling text-to-video when you don't have pre-recorded audio. Includes consent guardrails since driving someone's mouth with arbitrary audio is obviously dual-use. The docs are thorough on model trade-offs (Pro vs standard, cost vs quality, audio stem isolation) and all four routes ship with working bash invocations.
npx skills add https://github.com/agentspace-so/runcomfy-agent-skills --skill lipsync