black-forest-labs / flux-schnell

black-forest-labs/flux-schnell

A 12-billion parameter rectified flow transformer capable of generating images from text descriptions. Fast, 4-step generation with state-of-the-art prompt adherence.

Ready v1.0 · 12.4M runs · 640k downloads · Apache 2.0
Run
Overview Runs API Schema README Examples Changelog
# readme

FLUX.1 [schnell]

FLUX.1 [schnell] is a 12 billion parameter rectified flow transformer capable of generating images from text descriptions. It is trained to generate images of exceptional quality and prompt adherence, competitive with closed-source alternatives. The schnell variant is optimized for fast, 4-step inference under an Apache 2.0 license.

This model produces output suitable for commercial use. Pass a text prompt and receive a 1024×1024 image in under a second on H100 hardware. See the API reference for details.

# input schema
input.json
{
  "prompt": {        // Required. text → string
    "type": "string",
    "description": "Text prompt for image generation"
  },
  "num_outputs": {  // Default 1
    "type": "integer",
    "default": 1,
    "maximum": 4
  },
  "aspect_ratio": {
    "type": "string",
    "default": "1:1",
    "options": ["1:1", "16:9", "9:16"]
  },
  "output_format": {
    "type": "string",
    "default": "webp"
  },
  "num_inference_steps": {
    "type": "integer",
    "default": 4,
    "maximum": 12
  }
}
# output schema
output.json
{
  "output": {
    "type": "array",
    "items": { "type": "string", "format": "uri" },
    "description": "Generated image URLs"
  },
  "seed": {
    "type": "integer",
    "description": "Seed used for reproducibility"
  }
}
# run via api
$ replicate run black-forest-labs/flux-schnell \ -i prompt="a neon cyberpunk city at night, rain" \ -i num_inference_steps=4 \ -i aspect_ratio="16:9"