usamaehsan/qwen-image-edit-2511-lab

Qwen-Image-Edit-2511 cost lab: FP8 + torch.compile + SageAttention, tunable steps/reference size

Public

1K runs

Run usamaehsan/qwen-image-edit-2511-lab 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
prompt	string		Edit instruction
image	array		Reference image(s) as data: URIs
width	integer	Min: 256.0 Max: 1536.0	Output width, /16
height	integer	Min: 256.0 Max: 1536.0	Output height, /16
num_inference_steps	integer	0 Max: 12.0	0 = backend default (fp8: 4, nunchaku: 8)
ref_size	integer	0 Max: 1536.0	Side of the square area the reference is resized to for the VAE (its tokens join the transformer sequence). 0 = output long side (exact_size: 1024). Stock: 1024
cond_size	integer	384 Min: 128.0 Max: 768.0	Side of the area the reference is resized to for the text encoder's vision tower. Stock: 384
attention	None	native	None
seed	integer		None
output_format	None	jpg	None
output_quality	integer	90 Min: 1.0 Max: 100.0	None
exact_size	boolean	False	Use width/height/ref_size/cond_size exactly as given instead of snapping to the compiled size set. Runs the uncompiled transformer, so it is slower per step but never compiles
build_cache	boolean	False	Maintenance only: compile the warm size set and return the inductor cache as a tar instead of an image

{
  "type": "object",
  "title": "Input",
  "required": [
    "prompt",
    "image"
  ],
  "properties": {
    "seed": {
      "type": "integer",
      "title": "Seed",
      "x-order": 8
    },
    "image": {
      "type": "array",
      "items": {
        "type": "string"
      },
      "title": "Image",
      "x-order": 1,
      "description": "Reference image(s) as data: URIs"
    },
    "width": {
      "type": "integer",
      "title": "Width",
      "maximum": 1536.0,
      "minimum": 256.0,
      "x-order": 2,
      "description": "Output width, /16"
    },
    "height": {
      "type": "integer",
      "title": "Height",
      "maximum": 1536.0,
      "minimum": 256.0,
      "x-order": 3,
      "description": "Output height, /16"
    },
    "prompt": {
      "type": "string",
      "title": "Prompt",
      "x-order": 0,
      "description": "Edit instruction"
    },
    "ref_size": {
      "type": "integer",
      "title": "Ref Size",
      "default": 0,
      "maximum": 1536.0,
      "minimum": 0.0,
      "x-order": 5,
      "description": "Side of the square area the reference is resized to for the VAE (its tokens join the transformer sequence). 0 = output long side (exact_size: 1024). Stock: 1024"
    },
    "attention": {
      "enum": [
        "native",
        "sage"
      ],
      "type": "string",
      "title": "attention",
      "description": "An enumeration.",
      "default": "native",
      "x-order": 7
    },
    "cond_size": {
      "type": "integer",
      "title": "Cond Size",
      "default": 384,
      "maximum": 768.0,
      "minimum": 128.0,
      "x-order": 6,
      "description": "Side of the area the reference is resized to for the text encoder's vision tower. Stock: 384"
    },
    "exact_size": {
      "type": "boolean",
      "title": "Exact Size",
      "default": false,
      "x-order": 11,
      "description": "Use width/height/ref_size/cond_size exactly as given instead of snapping to the compiled size set. Runs the uncompiled transformer, so it is slower per step but never compiles"
    },
    "build_cache": {
      "type": "boolean",
      "title": "Build Cache",
      "default": false,
      "x-order": 12,
      "description": "Maintenance only: compile the warm size set and return the inductor cache as a tar instead of an image"
    },
    "output_format": {
      "enum": [
        "webp",
        "jpg",
        "png"
      ],
      "type": "string",
      "title": "output_format",
      "description": "An enumeration.",
      "default": "jpg",
      "x-order": 9
    },
    "output_quality": {
      "type": "integer",
      "title": "Output Quality",
      "default": 90,
      "maximum": 100.0,
      "minimum": 1.0,
      "x-order": 10
    },
    "num_inference_steps": {
      "type": "integer",
      "title": "Num Inference Steps",
      "default": 0,
      "maximum": 12.0,
      "minimum": 0.0,
      "x-order": 4,
      "description": "0 = backend default (fp8: 4, nunchaku: 8)"
    }
  }
}

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"
}