FLUX.3 Image Klein is available in the Diffusers library for image generation and editing
The FLUX.3 Image Klein model from Black Forest Labs is available in the Diffusers library with two variants of the weights. It supports text-to-image generation, editing with multiple reference images, and control over the placement of elements using bounding boxes.
Hugging Face has introduced an integration of the FLUX.3 Image Klein model from Black Forest Labs into the Diffusers library. Distilled and base variants of the weights are available. The model supports text-to-image generation, editing with multiple reference images, and composition control using bounding boxes, rectangular regions that specify where elements are placed. It can be tried in a public demo on Hugging Face.
The distilled variant uses 16 steps without CFG by default, while the base variant uses 50 steps with CFG. According to Hugging Face, the distilled variant generates a 1024 × 1024 image in approximately two seconds on an H100 GPU. Weights in NVFP4 and FP8 formats are also available for smaller GPUs and can be loaded using from_single_file.
The article includes instructions for running the model through ModularPipeline and examples of working with reference images and scene layouts. Region coordinates use a range of 0–1000, so they are independent of any specific resolution. The integration allows individual blocks to be replaced or added and includes scripts for LoRA training for both text-to-image generation and image editing.
Why it matters
When composing a scene, users can combine content from several reference images and specify the position of individual elements. Edits can therefore target specific image regions; the article also shows a prompt for moving an object while preserving the rest of the scene.
Two audiences, two different impacts
What this means
For individuals
Image creators can try generation in the public demo and use multiple reference images or specify the placement of elements when editing.
For a business
Development teams gain an integration for their own image applications, with customizable blocks and scripts for LoRA training.
DevelopmentCheck the original
Event sources
clearly official source · 1 publisher, 0 independent. We count feeds from the same owner only once.