irulkenzei/musetalk

Public
8 runs

Run irulkenzei/musetalk with an API

Use one of our client libraries to get started quickly. Clicking on a library will take you to the Playground tab where you can tweak different inputs, see the results, and copy the corresponding code to use in your own project.

Input schema

The fields you can use to run this model with an API. If you don't give a value for a field its default value will be used.

Field Type Default value Description
face
string
Source face -- a still image or a talking-head video. This identity gets lip-synced to 'audio'.
audio
string
Driving audio (wav/mp3/etc.) that the face's mouth will be synced to.
version
None
v15
MuseTalk model version. v15 is the recommended/higher-quality version.
bbox_shift
integer
0

Min: -9

Max: 9

Vertical face bounding-box shift in pixels (v1 only, ignored for v15). Positive = lower mouth region.
extra_margin
integer
10

Max: 40

Extra chin margin in pixels for the face crop (v15 only).
parsing_mode
None
jaw
Face mask blending mode used when compositing the generated mouth back onto the source frame (v15 only).
left_cheek_width
integer
90

Min: 20

Max: 160

Left cheek mask width in pixels (v15 only).
right_cheek_width
integer
90

Min: 20

Max: 160

Right cheek mask width in pixels (v15 only).
fps
integer
25

Min: 1

Max: 60

Output frame rate. Only used when 'face' is a still image -- a video 'face' input keeps its own source FPS.
audio_padding_length_left
integer
2

Max: 10

Left audio context padding, in Whisper feature frames.
audio_padding_length_right
integer
2

Max: 10

Right audio context padding, in Whisper feature frames.
batch_size
integer
8

Min: 1

Max: 32

Inference batch size. Lower this if you hit GPU out-of-memory.
use_float16
boolean
True
Run in float16 for lower VRAM usage and faster inference, at a small quality cost.

Output schema

The shape of the response you’ll get when you run this model with an API.

Schema
{
  "type": "string",
  "title": "Output",
  "format": "uri"
}