AiNews.com
Posts
FLUX.1 Tools Launch: Advanced Image Editing Models Released

FLUX.1 Tools Launch: Advanced Image Editing Models Released

Alicia Shapiro
November 22, 2024 • Estimated Reading Time: 11 minutes

A futuristic concept image of the FLUX.1 Tools suite, showcasing a sleek digital interface. Glowing icons represent tools for inpainting, outpainting, depth mapping, edge detection, and image variation. The interface is set against a gradient background of blue and purple, reflecting innovation and user-friendly design for advanced image editing and AI technology.

Image Source: ChatGPT-4o

FLUX.1 Tools Launch: Advanced Image Editing Models Released

Black Forest Labs (BFL) announced the release of FLUX.1 Tools, a suite of advanced models built on the base FLUX.1 text-to-image framework. These tools introduce enhanced control and steerability for editing and recreating both real and generated images. Available as open-access and API-powered tools, FLUX.1 Tools offer four distinct features:

FLUX.1 Fill: Provides state-of-the-art inpainting and outpainting for seamless image edits and expansions.
FLUX.1 Depth: Uses depth mapping for structural guidance, preserving an image’s core composition during edits.
FLUX.1 Canny: Employs canny edge detection to maintain structural integrity during text-guided transformations.
FLUX.1 Redux: Allows image variation and restyling, blending input images with text prompts to create new outputs.

Each tool will be part of the FLUX.1 [dev] model series as open-access versions and integrated into the BFL API as FLUX.1 [pro] variants. Additionally, the models will be accessible via partnerships with fal.ai, Replicate, Together.ai, Freepik, and krea.ai.

Key Features in Detail

Inpainting and Outpainting with FLUX.1 Fill

FLUX.1 Fill enables seamless image edits and expansions, delivering results that outperform competitors like Ideogram 2.0 and AlimamaCreative’s FLUX-Controlnet-Inpainting.

Inpainting: Enables natural, unnoticeable edits to existing images.

A three-panel image showcasing FLUX.1 Fill's capabilities. The first panel shows a man in a gray suit standing outdoors against a background of tall grass. The second panel depicts the same image with the suit masked in black. The final panel shows the man’s suit replaced with a casual denim jacket and tie, demonstrating FLUX.1 Fill's ability to modify clothing seamlessly.

Image Source: Black Forest Labs

Outpainting: Expands images beyond their original borders, maintaining visual harmony.

A two-panel image illustrating the outpainting feature of FLUX.1 Fill. The first panel highlights a close-up of a man’s eye, with directional arrows pointing outward to indicate the expansion of the image. The second panel shows the completed outpainting, revealing the man's full face in sharp detail, seamlessly extended from the original close-up.

Image Source: Black Forest Labs

According to benchmarks, FLUX.1 Fill [pro] outshines all competing models in both accuracy and efficiency, with the [dev] version surpassing proprietary solutions while offering faster inference.

A bar chart comparing the ELO scores of different inpainting models. The chart shows FLUX.1 Fill [pro] at the top with a score of 1094, followed by FLUX.1 Fill [dev] at 1040, Ideogram V2 at 1020, FLUX.1 IdeV CN Alimama Creative at 1016, and SD 1.5 Inpainting at 829. The chart uses a dark background with blue and purple bars to indicate performance levels.

Image Source: Black Forest Labs

Access Options for FLUX.1 Fill:

Full model weights and inference code are available under the Flux Dev License via Hugging Face and GitHub.

The professional-grade version, FLUX.1 Fill [pro], is accessible through the BFL API.

Structural Conditioning with FLUX.1 Canny and Depth

These models introduce structural conditioning, preserving core image details during transformations through depth mapping or edge detection, which is particularly effective for retexturing images.

FLUX.1 Depth: Utilizes depth maps to maintain a scene’s structure while enabling text-guided edits and retexturing.
FLUX.1 Canny: Focuses on preserving image edges for precise transformations.

A series of transformations showcasing FLUX.1 Depth and Canny capabilities. The top row features an image of animals in a hole, transformed through depth mapping into fantasy-themed unicorns and playful alien characters. The bottom row shows a photo of an owl transformed via canny edge detection into a glittered owl, a grayscale owl, and a vibrant owl in a fantasy setting with butterflies. Each panel demonstrates structural preservation while enabling creative reimaginings.

Image Source: Black Forest Labs

Benchmarks highlight FLUX.1 Depth’s superior performance in depth-aware tasks and FLUX.1 Canny’s precision in structural editing. Both [pro] and [dev] versions offer full functionality, with LoRA weights for easier integration in developer workflows.

Two side-by-side bar charts comparing ELO scores of depth and canny-based image editing models. The left chart shows FLUX.1 Depth [dev] scoring 1098 and FLUX.1 Depth [pro] scoring 1074, outperforming competitors like MidJourney Retexture (1059) and Jasper (838). The right chart highlights FLUX.1 Canny [pro] with a score of 1078 and FLUX.1 Canny [dev] with 1055, surpassing SD 1.5 CN Ilvasviel (971) and SDXL CN Diffusers (915). Both charts use a dark background with blue and purple bars to visualize performance.

Image Source: Black Forest Labs

Access Options for FLUX.1 Canny and Depth:

Full model weights are available on Hugging Face. Depth. Canny.
LoRA model weights are available on Hugging Face. Depth. Canny.
Inference code is on Github.
Professional versions are included in the BFL API.

Image Variation and Restyling with FLUX.1 Redux

FLUX.1 Redux is designed to refine input images through subtle variations or to restyle them via text prompts.

Image Variation: Generates alternative versions of input images with slight modifications.

A four-panel sequence demonstrating FLUX.1 Redux's ability to restyle food imagery. The first panel shows a loaf of bread, a cup of milk, and a block of butter on a white surface. The second panel highlights a new version of the bread with more defined scoring. The third and fourth panels showcase further variations, including changes in the bread’s texture and shape, while maintaining the same arrangement of milk and butter.

Image Source: Black Forest Labs

Restyling: Allows users to blend images with textual inputs for customized outputs, supported by FLUX1.1 [pro] Ultra for high-resolution 4-megapixel images and flexible aspect ratios.

A five-panel sequence showing a duck character transformed into various styles using FLUX.1 Redux. The transformations include: a fine glass sculpture on a shelf, a simple 3D model outdoors, a realistic duck on a desk, and an epic anime-style duck flying through clouds. Each panel highlights the versatility of the tool in adapting the character to vastly different creative contexts.

Image Source: Black Forest Labs

Benchmarks confirm FLUX.1 Redux’s state-of-the-art performance in generating diverse and high-quality image variations.

A bar chart comparing ELO scores for image variation models. FLUX.1 Redux [pro] leads with 1202, followed by FLUX.1 Redux [dev] at 1171, and FLUX.1 Redux [schnell] at 1000. Competing models, SD 1.5 Image Variations and SDXL Image Variations, trail with scores of 883 and 743, respectively. The chart uses a dark background with blue and purple bars to depict performance differences.

Image Source: Black Forest Labs

Access Options for FLUX.1 Redux:

Full model weights and inference code are available under the Flux Dev License on Hugging Face and GitHub.

The advanced FLUX1.1 [pro] Ultra version is offered through the BFL API.

Commitment to Accessibility and Innovation

The release of FLUX.1 Tools represents a transformative step in the AI creative industry, bridging the gap between cutting-edge innovation and accessibility. By offering both open-weight models for researchers and premium capabilities through its API, FLUX.1 Tools democratize advanced image editing technologies. This dual approach empowers developers, businesses, and creators to push the boundaries of what’s possible with AI, from enhancing workflows to delivering unparalleled creative outputs.

FLUX.1 Tools also encourage collaboration across platforms, thanks to partnerships with services like Replicate, Freepik, and Together.ai, expanding the ecosystem of tools available for creators worldwide. By emphasizing accessibility without sacrificing performance, FLUX.1 Tools pave the way for new possibilities in AI-driven creativity.

What This Means

The introduction of FLUX.1 Tools highlights the rapid evolution of AI technology within the creative industry. These tools offer a glimpse into the future of image editing, where AI can seamlessly collaborate with human creators to unlock new levels of precision, creativity, and efficiency.

For professionals in industries like design, media, and advertising, tools like FLUX.1 Fill and FLUX.1 Redux streamline workflows by offering state-of-the-art inpainting, structural guidance, and image variation capabilities. At the same time, open-weight access fosters innovation within the research community, encouraging experimentation and advancements in AI image generation.

As the FLUX ecosystem expands, it will likely set new standards for AI-powered creativity, reducing barriers for individual creators and small teams to access high-quality tools. This democratization of technology signals a future where AI becomes an integral and accessible part of creative endeavors.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.