You're looking at a specific version of this model. Jump to the model overview.
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 |
|---|---|---|---|
| media |
string
|
Video (mp4/mov) or image (png/jpg/webp) to restore.
|
|
| cfg_scale |
number
|
1
Max: 15 |
Classifier-free guidance scale (higher = stronger restoration).
|
| sample_steps |
integer
|
1
Min: 1 Max: 4 |
Sampling steps (1 = fast one-step mode).
|
| sp_size |
integer
|
1
Min: 1 Max: 4 |
Sequence-parallel shard heuristic (single-GPU build only accepts 1).
|
| fps |
integer
|
24
Min: 1 Max: 120 |
Frames-per-second for video outputs.
|
| seed |
integer
|
-1
|
Random seed. Use -1 for random seed each call.
|
| output_format |
None
|
png
|
Image output format (only used for image inputs).
|
| output_quality |
integer
|
90
Min: 10 Max: 100 |
Image quality for lossy formats (jpg/webp).
|
| apply_color_fix |
boolean
|
False
|
Apply optional wavelet color correction (matches official demo).
|
| model_variant |
None
|
7b
|
Model size to run.
|
| resolution |
integer
|
1080
Min: 16 Max: 16384 |
Target resolution for the shortest edge in pixels (default: 1080).
Automatically maintains aspect ratio of input.
Even values required for optimal processing.
|
| max_resolution |
integer
|
3840
Max: 16384 |
Maximum resolution limit for any dimension (default: 0, no limit).
If any edge exceeds this after applying resolution,
both dimensions are scaled down proportionally.
Useful to prevent excessive VRAM usage on extreme aspect ratios.
|
Output schema
The shape of the response you’ll get when you run this model with an API.
{'format': 'uri', 'title': 'Output', 'type': 'string'}