AI Briefingflux-3-image-klein: Integrated text and image editing with 5.6B parameters
black-forest-labs/flux-3-image-klein
About the project
It performs text generation, single-image editing, and multi-image synthesis in a single model. By combining a 5.6B-parameter image transformer with the Qwen3-VL-4B-Instruct text encoder, it takes a prompt and up to 10 reference images as input and outputs a result image.
Bounding boxes can be specified within the prompt using a coordinate grid ranging from 0 to 1000. This allows control over element position and size, enabling fine-grained edits such as color changes, movement, replacement, and deletion. It also supports text rendering within generated images.
Various quantization variants, including BF16, FP8, and NVFP4, are provided to allow selection based on hardware environments. In the reference implementation, generating images at 1024x1024 resolution on a B200 GPU uses 31GB of memory for BF16 and 26GB for FP8. Under the FLUX Kommunity license, non-commercial use and commercial output use under specific conditions are permitted.
black-forest-labs/flux-3-image-klein
The original page has no description.
image-to-image
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.