FLUX Tools

Usage

FLUX Fill is used with the following specific configuration:

• Use base model AIR ID runware:102@1.
• Provide seedImage and maskImage as you would for standard inpainting.
• Unlike regular inpainting models, FLUX Fill does not support the maskMargin parameter for zoomed/detailed inpainting.
• The strength parameter is also not compatible with this model, the balance between existing and new content is controlled entirely through prompting.

{
  "taskType": "imageInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "model": "runware:102@1",
  "positivePrompt": "a blue denim jacket",
  "seedImage": "59a2edc2-45e6-429f-be5f-7ded59b92046",
  "maskImage": "b6a06b3b-ce32-4884-ad93-c5eca7937ba0",
  "width": 1024,
  "height": 1024,
  "steps": 30
}

Use cases

FLUX Fill excels at:

• Object Replacement: Replace specific objects in images while maintaining lighting and context.
• Background Modification: Change backgrounds while preserving the main subject.
• Content Extension: Expand images beyond their boundaries through outpainting.
• Detail Enhancement: Add or refine details within specific areas of an image.

Playground

Usage

• Use base model AIR ID runware:104@1 for FLUX Canny or runware:103@1 for FLUX Depth.
• Provide the structural guide image (edge map or depth map) directly in the seedImage parameter.
• There is no weight parameter for FLUX Canny/Depth. The strength of the conditioning can't be controlled.

{
  "taskType": "imageInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "model": "runware:103@1",
  "positivePrompt": "a watercolor painting of a forest",
  "seedImage": "59a2edc2-45e6-429f-be5f-7ded59b92046",
  "width": 1024,
  "height": 1024,
  "steps": 30
}

Preparing guide images

You can use our ControlNet preprocessing tools to generate appropriate edge or depth maps:

• For FLUX Canny: Use the canny preprocessor type.
• For FLUX Depth: Use the depth preprocessor type.

{
  "taskType": "imageControlNetPreProcess",
  "taskUUID": "3303f1be-b3dc-41a2-94df-ead00498db57",
  "inputImage": "ff1d9a0b-b80f-4665-ae07-8055b99f4aea",
  "preProcessorType": "canny"
}

Use Cases

FLUX Canny/Depth excel at:

• Style Transfer: Transform image styles while maintaining structural composition and spatial relationships.
• Content Preservation: Generate new images that follow the exact structure of reference images.
• Scene Retexturing: Modify materials and textures while preserving object shapes and positions.
• Artistic Reinterpretation: Create artistic variants of photos with consistent structure but creative styling.
• Consistent Series Generation: Produce multiple variations with identical structural elements but different details.

Playground

Usage

FLUX Redux requires a different approach than the other FLUX tools:

• Use iP-Adapter model AIR ID runware:105@1.
• Provide the input image in the guideImage parameter inside the ipAdapters object.
• Use a FLUX base model (typically FLUX dev model - runware:101@1) as the base model. Other FLUX models can be used as well.
• There is no weight parameter for FLUX Redux. The variation level and guidance can be controlled through the positivePrompt parameter.

[
  {
    "taskType": "imageInference",
    "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
    "model": "runware:101@1",
    "positivePrompt": "elegant portrait, detailed features",
    "width": 1024,
    "height": 1024,
    "steps": 30,
    "ipAdapters": [
      {
        "guideImage": "59a2edc2-45e6-429f-be5f-7ded59b92046",
        "model": "runware:105@1"
      }
    ]
  }
]

Use cases

FLUX Redux excels at:

• Image Variations: Generate subtle alternatives of an input image while preserving key visual elements.
• Style Adaptation: Modify the artistic style of an image while maintaining subject recognition.
• Visual Concept Mixing: Combine visual concepts from the input image with new elements specified in the prompt.
• Subject Preservation: Ensure specific subjects or elements remain recognizable across style transformations.

Playground