ReferenceAsk
Back to Remy
Remy Reference/models/InfiniteTalk
W

InfiniteTalk

InfiniteTalk, a lipsync model from WaveSpeed.
lipsyncModelOverride: { model: "wavespeed-ai-infinitetalk" }
WaveSpeed·Released 2025
Pricing
List price
$0.03-$0.06/second
Per generation
$0.00010
Specifications
Context window
2,000 tokens
Aspect ratios
2
Released
2025
Model string
wavespeed-ai/infinitetalk
Parameters
Model options, passed in lipsyncModelOverride.config.
ParameterTypeDefaultRange / options
$modeSelectimageimagevideomultivideo-multi
image
Portrait to animate into a talking video.
Image URL——
video
Video to re-dub, keeping its motion and replacing the lips.
Video URL——
audio
Speech to lip sync to. The output runs as long as the audio.
Audio URL——
left_audio
Speech for the person on the left.
Audio URL——
right_audio
Speech for the person on the right.
Audio URL——
order
Whether the two speakers talk at the same time or take turns. Defaults to the same time.
Select—meanwhileleft_rightright_left
mask_image
Optional mask marking which face to animate, for an image or video with more than one person.
Image URL——
prompt
Optional description of the expression and movement.
Prompt——
resolution
The resolution of the output video.
Select720p480p720p
seedSeed-1—
Use this model
await mindstudio.generateLipsync({
  lipsyncModelOverride: {
    model: "wavespeed-ai-infinitetalk",
    config: {
      image: "...",
      video: "...",
      audio: "...",
      left_audio: "...",
      right_audio: "...",
      order: "meanwhile",
      resolution: "720p",
    },
  },
});