Black Forest Labs just brought on legendary filmmaker Martin Scorsese as an advisor to guide its visual intelligence development with a focus on human craft and taste. The announcement came alongside a working storyboarding session where Scorsese tested its FLUX image generation model.
Researchers just introduced SEGA, a training-free method that lets diffusion transformers generate coherent images far beyond their original training resolution. By dynamically scaling RoPE embeddings according to spatial-frequency energy at each denoising step, it balances global layout with fine textures. It means existing models like Flux can output 4096×4096 visuals without costly fine-tuning.
ByteDance researchers introduced a spatial reward modeling framework alongside SpatialScore, fixing a stubborn flaw where image generators mix up object positions. By training on over 80,000 spatial preference pairs, the method helps models like Flux follow complex layout prompts. The team plans to open-source the dataset and weights soon.
Meta just signed a two-year deal worth $140 million with German startup Black Forest Labs to license its Flux image generation technology. The contract is backloaded, with Meta paying $35 million in year one and $105 million in year two. Why it matters: Following its recent Midjourney pact, Meta is pouring serious cash into leading visual labs to supercharge its creative tools.
lmarena.ai just launched Image Arena, bringing blind side-by-side battles to AI image generation. Users enter a prompt to get two anonymous outputs from models like FLUX, Stable Diffusion, DALL-E, Recraft, and Ideogram, then vote for the winner ahead of an upcoming leaderboard.
Krea just rolled out Flux image prompts, allowing creators to condition their image generations using reference pictures. The added image-to-image steering is completely free for everyone to use.
Black Forest Labs just launched Flux 1.1 Pro, an updated image generation model claiming six times faster generation speeds alongside improved visual quality and prompt adherence. The startup also launched a beta API, letting developers integrate the Flux model suite into their apps via its platform or partners like Replicate.
Hobbyists just figured out how to inject custom typography directly into AI-generated images using the new Flux model. By training custom LoRA adapters, users can now render specific typefaces onto objects like chalkboards or business cards with surprising precision. While text in AI art used to be pure gibberish, Flux is turning in-scene lettering into a controllable design tool.
