FLUX.1 : Le générateur d'images open-source de Black Forest Labs

Multi-Reference Support with up to 10 simultaneous reference images for unmatched character consistency. Introduced JSON-based control system for more precise scene manipulation and workflow integration. Enabled Generative Expand/Shrink for dynamic composition adjustments. Improved pose guidance for accurate object and character positioning. Major upgrade to image detail and photorealism, delivering near-real photographic quality across textures, lighting, and perspective. Improved text rendering, now supporting complex typography, UI mockups, and production-grade readable text. Better color accuracy with precise HEX-based color matching. Upgraded enhanced prompt following, improving responsiveness to structured and complex instructions. Expanded world knowledge, resulting in more coherent physics, lighting, and real-world spatial logic. Faster generation speeds with sub-10-second output. Support for 32K token text inputs for more detailed prompts. Increased reliability across both low-res drafts and 4MP high-resolution outputs.

1. Speed and Efficiency - Six Times Faster: The model generates images six times more quickly, enhancing workflow efficiency. 2. Image Quality - Superior Photo-Realism: Improved clarity and texture rendering, with support for up to 2K ultra-high resolution images. 3. Enhanced Prompt Adherence and Diversity - Better Understanding of Prompts: Increased accuracy in interpreting complex prompts, leading to more faithful representations. Visual Diversity: Offers a wider range of stylistic outputs for creative exploration. 4. Reduced Latency - Significant reductions in latency streamline the image generation process for real-time applications. These enhancements make FLUX 1.1 a top choice for high-performance generative AI tools.

Initial release of FLUX.