FLUX 3 Image Klein Prompt Upsampler
The Klein Prompt Upsampler is a 9B vision-language model fine-tuned from Qwen3.5-9B. It turns a short image request into a detailed caption for the image generator.
Model details
The component handles editing, text-to-image and layout requests. BFL's reference pipeline loads it by default. It is a separate model from both Klein's image transformer and its Qwen3-VL text encoder.FLUX 3 Image Klein Prompt Upsampler official documentation
Practical use
Practical guidance. Expose the expanded prompt during development so added assumptions are easy to find. Evaluate whether the caption preserves the requested subject, identity and layout. Do not count this component as an additional image-generation model; its output is text.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | black-forest-labs/flux-3-image-klein-pup |
| Weight license | Apache 2.0FLUX 3 Image Klein Prompt Upsampler official documentation |
| Inputs | Request, up to ten reference images and optional layout boxes.FLUX 3 Image Klein Prompt Upsampler official documentation |
| Output | One expanded text prompt.FLUX 3 Image Klein Prompt Upsampler official documentation |
Related models
- FLUX 3 Image Klein
Black Forest Labs · FLUX 3
Distilled 5.6-billion-parameter image transformer for generation, editing and compositions with up to ten references.
Launch date not confirmedAPI · Downloadable · Provider product Undistilled 5.6-billion-parameter image model for generation, editing, fine-tuning and LoRA training.
Launch date not confirmedDownloadable