inference · 10.4s

Flux Fill Pro

Image·flux·by Black Forest Labs

Flux Fill Pro Model fills missing or unwanted areas in images seamlessly, ensuring realistic outcomes

Try it now →

API reference

Runtime (p50): 9s
Estimated price: $0.05

Call the API

prediction.sh

curl -X POST \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -H "Content-Type: application/json" \
  --data '{
    "model": "flux-fill-pro",
    "version": "0.0.1",
    "input": {
        "mask": "https://storage.googleapis.com/magicpoint/inputs/flux-fill-input-mask.webp",
        "image": "https://storage.googleapis.com/magicpoint/inputs/flux-fill-image-input.webp",
        "steps": 50,
        "prompt": "movie poster says \"FLUX FILL\"",
        "guidance": 3,
        "output_format": "jpg",
        "safety_tolerance": 2,
        "prompt_upsampling": false
    },
    "webhook_url": ""
}' \
  https://api.eachlabs.ai/v1/prediction/

Documentation8 sections

Overview
flux-fill-pro — Image-to-Image AI Model

flux-fill-pro, developed by Black Forest Labs as part of the flux family, excels at filling missing or unwanted areas in images seamlessly, delivering photorealistic results with production-grade quality. This image-to-image AI model solves the challenge of inpainting and outpainting by leveraging multi-reference editing—up to 10 input images for precise control over composition, poses, and styles—making it ideal for professional workflows like product visualization and creative editing. Users searching for "AI image editor API" or "edit images with AI" find flux-fill-pro's high-fidelity outputs stand out, supporting resolutions up to 4MP in any aspect ratio for sharp, detailed enhancements.
Capabilities
Image Inpainting: Modify or fill specific regions in an image while maintaining contextual consistency.
Creative Edits: Generate variations of an image based on textual descriptions, blending creativity with realism with Flux Fill Pro.
Refinement: Adjust existing images with enhanced textures, colors, and details.
Use cases
Use Cases for flux-fill-pro

E-commerce marketers use flux-fill-pro to remove backgrounds from product shots and fill with custom scenes, like placing sneakers on a urban street at dusk using multi-reference inputs for realistic shadows and textures—streamlining "AI photo editing for e-commerce" without photoshoots.

Graphic designers leverage its text rendering prowess for inpainting legible typography into layouts; input an image with a blank banner and prompt "Add the text 'Black Forest Labs' below the main headline in elegant serif typography, centered", yielding clean, accurate in-image text that outperforms competitors.

Developers building apps integrate the flux-fill-pro API for automated workflows, such as expanding user-uploaded photos with contextual elements via 8+ references, ideal for "image to image AI model" apps handling diverse aspect ratios up to 4MP.

Content creators repair damaged images or extend compositions, feeding a portrait plus environment references to fill edges seamlessly, preserving photorealistic details like skin textures for social media-ready edits.
Tips & tricks
How to Use flux-fill-pro on Eachlabs

Access flux-fill-pro through Eachlabs Playground for instant testing—upload your base image, add up to 10 references, provide a text prompt specifying fill areas (e.g., "fill the empty sky with dramatic clouds matching the sunset hex #FF4500"), and select resolutions up to 4MP. Integrate via API or SDK for scalable apps, with outputs in standard formats delivering production-quality inpaints efficiently.
---
Technical spec
What Sets flux-fill-pro Apart

flux-fill-pro distinguishes itself in the image-to-image AI model landscape through its support for up to 10 reference images in the playground (8 via API), enabling complex edits that maintain character consistency and style transfer across multiple sources—far beyond typical single-image inpainting tools. This capability allows developers to build robust "automated image editing API" solutions for e-commerce, where product photos can be seamlessly composited with backgrounds or elements without artifacts.

Unlike generic editors, flux-fill-pro offers advanced controls like pose guidance, hex color matching, and structured prompting, ensuring edits align precisely with creative intent while producing up to 4MP outputs from 64x64 minimum inputs. These features empower "Black Forest Labs image-to-image" users to achieve professional results at scale, with pro-grade performance optimized for production without sacrificing detail or speed.
- Multi-reference editing (up to 10 images): References multiple sources for consistent identity and style in fills, enabling intricate scene expansions.
- 4MP resolution support: Delivers high-detail inpainting in any aspect ratio, perfect for print-ready or web-optimized visuals.
- Precise controls (pose, color, grounding): Matches hex colors and poses exactly, reducing iterations in professional pipelines.
Things to be aware of
Experiment with steps to balance speed and detail.
Test different guidance values to explore the creative flexibility of your prompt.
Use masks creatively to isolate specific areas for focused modifications.
Combine prompts with high-resolution inputs to push the boundaries of inpainting capabilities.
Key considerations
Model Constraints: While Flux Fill Pro is powerful, it relies heavily on the quality and accuracy of inputs. Poorly designed masks or vague prompts may lead to suboptimal results.
Safety Tolerance: Higher safety tolerance values may limit creativity, while lower values allow more experimental outputs.
Parameter Limits: Inputs like steps and guidance have predefined ranges. Exceeding these may result in errors or inefficiencies.

Legal Information
By using this model, you agree to:
- Black Forest Labs API agreement
- Black Forest Labs Terms of Service
- Black Forest Labs Privacy Policy
Limitations
Flux Fill Pro may struggle with extremely vague prompts or poorly defined masks.
Outputs depend heavily on the accuracy of input parameters, requiring careful calibration.
Overly complex prompts may lead to unexpected or less cohesive results.

Output Format: JPG,PNG