Readme
Chamelaion Lip-Sync
Lip-syncs a video to any audio track. Give it a source video and an audio file, and it returns a new video where the on-screen speaker’s mouth movements match the audio — ideal for dubbing, localization, and translating content into other languages while keeping it natural.
This model is a thin wrapper around the Chamelaion lip-sync API; all generation happens on Chamelaion’s service.
Inputs
| Input | Type | Description |
|---|---|---|
video_url |
string (required) | Public URL of the source video (MP4). |
audio_url |
string (required) | Public URL of the audio track to lip-sync to (WAV/MP3). |
disable_active_speaker_detection |
boolean (default false) |
Active-speaker detection finds which face is talking and syncs only that person — needed when multiple faces are visible. Turn this on only for single-speaker videos for faster processing. |
Both URLs must be publicly reachable — the service fetches them directly (no file upload).
Output
A lip-synced MP4 video, hosted on Replicate. Generation is asynchronous; a prediction stays running while the job is processed and returns the finished video when it’s ready.
Example
import replicate
output = replicate.run(
"<owner>/lipsync",
input={
"video_url": "https://example.com/source.mp4",
"audio_url": "https://example.com/voice_es.mp3",
# "disable_active_speaker_detection": True, # single-speaker videos only
},
)
print(output) # URL of the lip-synced video
Intended use
- Dubbing and localizing videos into other languages with natural, matched lip movement.
- Re-voicing existing footage (education, marketing, accessibility, personalized messaging).
Limitations
- Best results with clear, front-facing footage and clean audio.
- Video and audio must be accessible via public URLs.
- Multi-speaker scenes rely on active-speaker detection; disabling it on such videos can sync the wrong face.
Ethical considerations
This model alters a person’s mouth movements to match audio they may not have spoken. Only use it on footage you own or have permission to modify, and with the consent of the people shown. Do not use it to impersonate, deceive, or create misleading content about real people. Consider disclosing when a video has been altered. You are responsible for complying with applicable laws and platform policies.