usamaehsan/ernie-image-turbo-fast

ERNIE-Image-Turbo tuned for cost: FP8 + torch.compile + lean forward, 1-8 steps, optional hires refine

Public

52 runs

Run usamaehsan/ernie-image-turbo-fast 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		Text prompt
width	integer	1024 Min: 256 Max: 2048	Rounded down to a multiple of 16
height	integer	1024 Min: 256 Max: 2048	Rounded down to a multiple of 16
num_inference_steps	integer	2 Min: 1 Max: 8	Denoising steps (the draft's steps when hires_from is set)
sigma_shift	number	4 Min: 1 Max: 12	Flow-matching timestep shift; the model card uses 4
hires_from	integer	0 Max: 2048	0 = off. Otherwise draft at this long side, upscale, then refine at full size
hires_steps	integer	1 Min: 1 Max: 4	Full-size refine steps when hires_from is set
hires_strength	number	0.35 Min: 0.05 Max: 1	Noise level the refine restarts from (before shift)
hires_mode	None	latent	Upscale the draft in latent space or through a VAE decode/encode
seed	integer		Random seed
output_format	None	jpg	None
output_quality	integer	90 Min: 1 Max: 100	None
eager	boolean	False	Debug: skip torch.compile (compare speed)
dump_compile_cache	boolean	False	Debug: return the torch.compile cache instead of the image

{
  "type": "object",
  "title": "Input",
  "required": [
    "prompt"
  ],
  "properties": {
    "seed": {
      "type": "integer",
      "title": "Seed",
      "x-order": 9,
      "description": "Random seed"
    },
    "eager": {
      "type": "boolean",
      "title": "Eager",
      "default": false,
      "x-order": 12,
      "description": "Debug: skip torch.compile (compare speed)"
    },
    "width": {
      "type": "integer",
      "title": "Width",
      "default": 1024,
      "maximum": 2048,
      "minimum": 256,
      "x-order": 1,
      "description": "Rounded down to a multiple of 16"
    },
    "height": {
      "type": "integer",
      "title": "Height",
      "default": 1024,
      "maximum": 2048,
      "minimum": 256,
      "x-order": 2,
      "description": "Rounded down to a multiple of 16"
    },
    "prompt": {
      "type": "string",
      "title": "Prompt",
      "x-order": 0,
      "description": "Text prompt"
    },
    "hires_from": {
      "type": "integer",
      "title": "Hires From",
      "default": 0,
      "maximum": 2048,
      "minimum": 0,
      "x-order": 5,
      "description": "0 = off. Otherwise draft at this long side, upscale, then refine at full size"
    },
    "hires_mode": {
      "enum": [
        "latent",
        "pixel"
      ],
      "type": "string",
      "title": "hires_mode",
      "description": "Upscale the draft in latent space or through a VAE decode/encode",
      "default": "latent",
      "x-order": 8
    },
    "hires_steps": {
      "type": "integer",
      "title": "Hires Steps",
      "default": 1,
      "maximum": 4,
      "minimum": 1,
      "x-order": 6,
      "description": "Full-size refine steps when hires_from is set"
    },
    "sigma_shift": {
      "type": "number",
      "title": "Sigma Shift",
      "default": 4,
      "maximum": 12,
      "minimum": 1,
      "x-order": 4,
      "description": "Flow-matching timestep shift; the model card uses 4"
    },
    "output_format": {
      "enum": [
        "jpg",
        "png",
        "webp"
      ],
      "type": "string",
      "title": "output_format",
      "description": "An enumeration.",
      "default": "jpg",
      "x-order": 10
    },
    "hires_strength": {
      "type": "number",
      "title": "Hires Strength",
      "default": 0.35,
      "maximum": 1,
      "minimum": 0.05,
      "x-order": 7,
      "description": "Noise level the refine restarts from (before shift)"
    },
    "output_quality": {
      "type": "integer",
      "title": "Output Quality",
      "default": 90,
      "maximum": 100,
      "minimum": 1,
      "x-order": 11
    },
    "dump_compile_cache": {
      "type": "boolean",
      "title": "Dump Compile Cache",
      "default": false,
      "x-order": 13,
      "description": "Debug: return the torch.compile cache instead of the image"
    },
    "num_inference_steps": {
      "type": "integer",
      "title": "Num Inference Steps",
      "default": 2,
      "maximum": 8,
      "minimum": 1,
      "x-order": 3,
      "description": "Denoising steps (the draft's steps when hires_from is set)"
    }
  }
}

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"
}