You're looking at a specific version of this model. Jump to the model overview.
usamaehsan /qwen-image-4bit:5190bd4f
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 |
|---|---|---|---|
| image |
array
|
Input image(s). Provide a single image for editing, or multiple images (2-4) for composition tasks
|
|
| prompt |
string
|
Remove the background
|
Text prompt describing the edit or composition to perform
|
| negative_prompt |
string
|
|
Negative prompt (things to avoid in the output)
|
| width |
integer
|
0
Max: 2048 |
Width of output image. Set to 0 to use input image width
|
| height |
integer
|
0
Max: 2048 |
Height of output image. Set to 0 to use input image height
|
| num_inference_steps |
integer
|
20
Min: 1 Max: 50 |
Number of denoising steps. More steps = higher quality but slower. Use 40+ for compositions
|
| guidance_scale |
number
|
1
Max: 20 |
Guidance scale (for multi-image composition)
|
| true_cfg_scale |
number
|
2
Min: 1 Max: 20 |
True guidance scale for the edit. Higher values = stronger adherence to prompt. Use 1.0 to disable CFG. Recommended: 2.0-4.0 for edits
|
| num_images_per_prompt |
integer
|
1
Min: 1 Max: 4 |
Number of images to generate per prompt
|
| lora_mode |
None
|
none
|
Lightning LoRA mode for faster inference. 'none' = standard quality (20-40 steps), 'lightning-4steps' = ultra-fast (4 steps), 'lightning-8steps' = fast (8 steps)
|
| seed |
integer
|
Random seed. Leave blank to randomize
|
Output schema
The shape of the response you’ll get when you run this model with an API.
Schema
{'format': 'uri', 'title': 'Output', 'type': 'string'}