Black Forest Labs

Introducing FLUX.2 [klein]. Blazing fast. Beautiful. Generate stunning images in under a second while maintaining exceptional quality. Great for fast editing, changing styles, and developing ideas from 0 → 1. Available via API, or run it locally - Klein 4B under Apache 2.0, Klein 9B as open weights. Try it for free in our demo app (link in the thread).

FLUX automatically removes related details like shadows.

Good at removing text while preserving design.

New Release: Erase Remove objects, erase text, and clean up details in any image. FLUX is now trained to erase and reconstruct as one task at the model level. We noticed this approach leads to images that stay clean and consistent. Give it an image + a simple “erase mask” over what you want gone, and it reconstructs the background - shadows and missed bits included - while keeping the rest intact. Try the demo: flux-tools.bfl.ai/erase Available on the BFL API: docs.bfl.ai/flux_erase

Try it for free: mcp.bfl.ai

The official FLUX MCP. Create images inside Claude, Cursor, Hermes, and more. Your agent automatically routes your request to the right FLUX model with expert-level prompting. From Klein for sub-second drafts to Pro for production-quality assets. Play with website visuals, ad creatives, social content, product renders, concept art, and (most importantly) memes. Link in the thread.

New release: Outpainting (expand any image to any aspect ratio). Most outpainting struggles at boundaries: light breaks, textures drift, seams appear. FLUX Outpainting fixes this at the model level. Give it an image + canvas geometry and get a coherent scene extension with preserved structure and continuity. Available via the BFL API: docs.bfl.ai/flux_2/flux_ou… Try the demo: flux-tools.bfl.ai/outpainting

Our cofounder @iamtimdavis built an AI storybook app using @BlackForestLabs' FLUX2 and @googlegemma 4 on Modular Cloud. Pick a character, make choices, and the story branches endlessly, with every page written and illustrated in real time. Tim has spent his career obsessing over inference latency, first at Google, now at Modular. Building something his kids use settled it: in a real-time generative app, the inference platform determines the experience as much as the model. The numbers back that up. From 24 hours of production traffic: first prose in 420ms, a full illustration in under 6 seconds, 85% of page turns in 48ms. Create your own story with Inkwell and share it. We're sending swag to our favorites: inkwell.modular.com

Self flow research: bfl.ai/research/self-…

Full video: youtube.com/watch?v=x8Yb4R…

The next generation of models won't just generate images - they'll understand worlds, motion, interaction, and action. We've been building toward this for a while. Visual intelligence is becoming real-time. @stephenbtl spoke about where we're headed at @aiDotEngineer:

If this kind of work resonates, we’re hiring: bfl.ai/careers

Full post: bfl.ai/blog/how-we-ge…

Meet @bfl_ml's AI Solutions team. Here's how our forward deployed team works with customers: from prompt design and inference pipelines, to fine-tuning, evaluation, and infrastructure.

Apply here: bfl.ai/creator-program

Be the first to get access to upcoming FLUX Models. We’re opening up the first FLUX Creator Program. We’re looking for a small group of creators already making standout work with FLUX. You get: • Early access to our upcoming models • Amplification of your work via BFL’s channels • Lots of credits • Direct access to our engineering team This is our first cohort, so spaces are limited.

BFL x @Forbes AI 50 list A nod to our researchers and community pushing the pixel frontier, building open infrastructure for visual intelligence. Congrats to this year’s AI 50 class forbes.com/lists/ai50/

Full video ↓ youtu.be/TNxXs20yhMQ?si…

@andi_blatt joined @AnjneyMidha at @Stanford’s CS153 to talk about: • GANs → diffusion → FLUX • The leap to consistent, controllable visuals • Why “action-capable” visual systems matter

Full article: wired.com/story/black-fo…

@WIRED covered where we're headed. At BFL, we train models to be as general as possible when it comes to processing and understanding visual data. Joint training on video, images, and audio = a physics-level understanding of the real world. It should also result in better image gen and video. Adding action prediction lets you plug into a robot to map how humans learn to a machine. Content creation, video, physical AI are natural extensions of solving visual intelligence at the foundation.