PixVerse Lipsync: Face Video Lip Sync & Dubbing on Models and API | RunComfy

PixVerse / PixVerse Lipsync#

This model is built for a simple job: keep the identity, framing, and performance of an existing talking-head clip while replacing the spoken track. Upload a video with a clear face and a separate audio file, then receive a clip whose mouth motion follows the new dialogue.

That workflow suits dubbing, localization, and dialogue swaps when you already like the picture and only need the lips to catch up. On RunComfy you can run PixVerse Lipsync in the model UI or through an HTTP API after media upload.

Highlights#

• Audio-led mouth motion: PixVerse Lipsync drives lip shapes from your replacement track instead of inventing new speech from text alone.
• Keeps the source performance: Identity, lighting, and camera framing stay close to the original shot while the mouth is retimed.
• Two-input workflow: Provide only a video URL and an audio URL—no resolution mode or extra style knobs to tune.
• Dubbing-ready loop: Useful when you localize ads, replace lines, or swap languages without restaging talent.
• Single generation call: After uploads finish, one request produces the synced video output.

Parameters#

Pricing#

Billing follows the uploaded audio length, rounded up to the next whole second. The minimum billed length is 3 seconds, and the maximum is 600 seconds.

How to Use#

1. Open the PixVerse Lipsync page on RunComfy and choose the Video to Video template.
2. Upload or paste a video URL where the speaker's face is visible, well lit, and mostly unobstructed around the mouth.
3. Upload or paste the audio URL that should drive the new lip motion.
4. Confirm both fields are filled—the model expects these two inputs only.
5. Start generation and wait for the single video-to-video pass to finish.
6. Preview the result, then download or pass the output URL into your next edit step.
7. If mouth timing looks soft, trim noisy silence from the audio and retry with a cleaner take.
8. For API use, upload media first, then call generate with the same audio and video fields.

Prompt & Reference Tips#

• Face clarity: Prefer frontal or near-frontal shots so PixVerse Lipsync can lock onto the mouth reliably.
• Lighting: Even light on the face beats heavy shadows across the lips.
• Occlusions: Avoid hands, microphones, or hair covering the mouth for long stretches.
• Audio quality: Clean speech with limited background noise yields tighter sync than muddy tracks.
• Length planning: Short dialogue clips are easier to review; remember billing rounds up and floors at 3 seconds.
• Match intent: Use audio whose pacing fits the on-screen performance so the synced mouth feels natural.
• Stable camera: Handheld shake is fine in moderation, but extreme motion blur around the face can weaken alignment.

How PixVerse Lipsync compares to other models#

Based on publicly available information:

• Vs. text-driven talking-head tools: This path starts from real footage and a finished audio file, so it replaces lips on an existing performance instead of synthesizing a full avatar from a script.
• Vs. image-to-talking-video systems: Those tools usually animate a still portrait; PixVerse Lipsync expects a moving source video and focuses on mouth retiming.
• Vs. full generative video editors: Broader video models rewrite scenes from prompts; here the scope stays narrow—sync lips to audio while preserving the shot.
• When to pick it: Choose this model when the picture is already approved and only the spoken track must change.