FLUX.1 Tools

BlackForestLabs Dev Team thrilled to introduce FLUX.1 Tools, a powerful suite of features designed to bring greater control and precision to the FLUX.1 text-to-image model. This release enables users to modify and recreate real and AI-generated images with unparalleled ease and accuracy. The FLUX.1 Tools suite comprises four distinct features, available both as open-access models within the FLUX.1 [dev] series and as advanced options in the BFL API for FLUX.1 [pro] users:

• FLUX.1 Fill: Cutting-edge inpainting and outpainting tools that allow users to edit and expand images using a text description and binary mask.

• FLUX.1 Depth: Models trained to provide structural guidance by leveraging depth maps from input images and combining them with text prompts.

• FLUX.1 Canny: A feature enabling structural guidance using canny edges extracted from input images, paired with text prompts.

• FLUX.1 Redux: A versatile adapter that blends and recreates input images based on combined visual and textual inputs.

This release emphasizes our dual mission: to empower the research community with open-weight models and to provide best-in-class capabilities through our API.

Open Access and Partner Integrations

Each FLUX.1 Tool is available in the BFL API as a FLUX.1 [pro] variant, alongside guidance-distilled, open-access versions within the FLUX.1 [dev] series. The models are accessible via leading platforms such as fal.ai, Replicate, Together.ai, Freepik, and krea.ai, ensuring seamless integration into diverse workflows.

We’re excited about the impact these tools will have in expanding the vibrant Flux ecosystem, providing users with unprecedented creative possibilities.

FLUX.1 Fill: Advanced Inpainting and Outpainting

The FLUX.1 Fill tool redefines inpainting and outpainting, enabling precise edits and seamless image expansion based on textual descriptions. It outperforms existing tools like Ideogram 2.0 and popular open-source options, such as AlimamaCreative’s FLUX-Controlnet-Inpainting, by delivering edits that integrate naturally with the original image.

Stay tuned for detailed performance analyses of these tools and insights into their applications. We look forward to seeing how creators and researchers alike will leverage the new capabilities of FLUX.1 Tools.

FLUX.1 Fill [dev] is offered under the Flux Dev License, making it accessible to a wide audience for non-commercial use. Developers and researchers can take advantage of:

• Full model weights: Available on Hugging Face.

• Inference code: Accessible on GitHub.

For commercial and professional applications, FLUX.1 Fill [pro] is available through the BFL API, offering enhanced capabilities tailored for enterprise use.

Structural Conditioning with FLUX.1 Canny and Depth

FLUX.1 introduces Canny and Depth models for structural conditioning, delivering unmatched precision in text-guided image transformations. These tools use either canny edge detection or depth mapping to retain the original image's structural integrity while enabling stylistic or text-driven edits.

By preserving the image’s framework through edge or depth maps, users can retexture and modify images without compromising their fundamental composition. This makes the models especially effective for tasks such as:

Refining textures: Apply new textures while maintaining the original shape and form.

Enhancing realism: Retain realistic structures during stylistic adjustments.

These advanced features extend the creative possibilities of FLUX.1, offering robust tools for precise and controlled image editing.

Image Variation and Restyling with FLUX.1 Redux

FLUX.1 Redux is a powerful adapter designed for all FLUX.1 base models, enabling seamless image variation and restyling. This feature is ideal for refining or experimenting with existing visuals, offering:

• Slight Variations: Generate alternate versions of a given input image while preserving its core elements.

• Creative Refinement: Perfect an image by exploring subtle changes in style, texture, or composition.

By integrating text-based prompts with input images, FLUX.1 Redux ensures a balance between creative flexibility and fidelity to the original image. It opens new doors for iterative design workflows, whether for artistic exploration, advertising, or professional-grade image adjustments.

FLUX.1 Redux seamlessly integrates into complex workflows, unlocking advanced image restyling through prompts. By providing an input image and a text description, users can transform and restyle visuals effortlessly, enabling a unique blend of creativity and control.

This feature is accessible via the BFL API, offering professional-grade capabilities with the latest FLUX.1 [pro] Ultra model. The tool supports high-quality 4-megapixel outputs, ensuring stunning image results while maintaining flexible aspect ratios. Whether for artistic reinvention or refining visuals for commercial projects, FLUX.1 Redux delivers unparalleled precision and versatility.

BFL's benchmark demonstrates that FLUX.1 Redux achieves state-of-the-art performance in image variation.

Flux.1 Redux [dev] is available under the Flux Dev License