FLUX 3 Image Klein

Use it to generate images from text, edit reference images, and create multi-reference compositions.

Availability

Publicly accessible repository; accept conditions and share Hugging Face information to access files

Pricing model

Not stated

Deployment

Self-hosting; ComfyUI files provided

Privacy

Not established

Licence

FLUX Kommunity License v.1.0; non-commercial use, with a commercial-output exception for Qualifying Users subject to terms

• The repository includes BF16, FP8, and NVFP4 transformer variants. ^[1]

  This repository includes the BF16 transformer and FP8 and NVFP4 variants, together with the Qwen3-VL-4B-Instruct text encoder and image autoencoder.

• The model supports text-to-image generation, single-reference editing, and multi-reference generation. ^[2]

  Text-to-image, single-reference edits and multi-reference compositions run through the same model, with Qwen3-VL-4B-Instruct as the text encoder.

• It takes up to 10 reference images. ^{[2] [3]}

  It takes a prompt and up to 10 reference images and returns one image.

• Bounding boxes guide placement and scale but do not constrain pixels to a hard mask. ^[2]

  Boxes guide placement and scale; they do not constrain pixels to a hard mask.

• The repository provides distilled BF16, int8, and NVFP4 checkpoints for ComfyUI. ^[2]

  This repository holds the model files repackaged for ComfyUI : the distilled and base transformers in BF16, int8 and NVFP4, the text encoder in BF16 and int8, the prompt upsampler in BF16, int8 and w4a8, and the image autoencoder.

• The distilled checkpoint is described as suitable for generation, editing, and self-hosting. ^[2]

  generation and editing, self-hosting; the checkpoint behind the BFL API

• The model is intended to use a prompt upsampler; short prompts produce generic output. ^[2]

  Always run a prompt upsampler, ours or your own LLM: the model takes full scene captions (framing, subject, background, lighting), and a short prompt sent as typed gives generic output.