lipsync

Provide a source video URL and an audio URL, then run Sync Labs sync v2 via the runcomfy CLI with both URLs in the JSON input. The model replaces only the mouth region while preserving the original camera, lighting, and body motion.

Sync Labs sync v2 applies mouth-sync onto an existing video, preserving the original footage. OmniHuman generates a new talking-head video from a single portrait photo plus audio, making it the default for avatar-style content rather than editing existing footage.

Yes. Kling lipsync text-to-video and HappyHorse 1.0 generate speech audio in-pass from a written script and sync it to the resulting video. The tradeoff is that the audio is regenerated each call, so you cannot lock the mouth to a specific voiceover.

Drift usually comes from a significant duration mismatch between the audio and video. Trim the audio or extend the video so lengths match, and use clean voiceover audio without a music bed, since audio quality directly drives mouth quality.

Only with consent from both parties. The skill instructs agents to refuse requests targeting real public figures without consent or aiming at defamatory or sexually explicit synthetic media, and operators must hold rights to both the face and the voice.