Sync Lipsync 2

Sync Lipsync 2 re-times the mouth in an existing video to a supplied recording. It keeps the framing and body performance. Attach one video and one audio track, with a clearly visible human-like face throughout. Kuro supports sections up to 30 minutes.

For long talks, loop a saved performance bank to the recording’s length first. Repeated motion will not follow the meaning of the words; use fresh avatar takes for the opening and closing. The Replicate route charges about $0.05 per second. A portrait alone belongs in an avatar model instead.

Keep hands and props away from the mouth, and avoid abrupt head turns or cuts to a different face. The audio sets the final length: a shorter base repeats, while a longer video is trimmed to the recording. Preview a short section with the same framing before committing a long narration.

Describe what you want in your own words and Kuro writes the prompt for you. If you would rather write it yourself, this is what Sync Lipsync 2 responds to.

Sync Lipsync 2 dubs one continuous recording over existing footage: attach one MP4 performance and one voice-over. It takes no prompt, because all it changes is the mouth.

Keep the performance and the recording the same length. A saved performance looped to the recording's length is the usual setup; the provider's loop mode repeats or cuts the footage to fit the audio.

Repeated base motion does not follow the meaning of the words. Use freshly generated hero takes for the hook and the conclusion, and let the loop carry the middle.

What it can't do

• Both a video and a recording are required, up to thirty minutes in one section.
• The output follows the source's framing. Keep one clearly visible, unobstructed, human-like face in frame; animated faces work less reliably.
• No image input. Make a performance from a portrait with an avatar model first.
• Pricing follows duration, and long sections take substantially longer than short avatar takes.

Use something else when…

• You have only a portrait and a short recording: OmniHuman 1.5.
• It is a short existing take: Kling Lip Sync.