A Black Forest Labs kiadta a FLUX.3 Image Klein képgeneráló modellt, amely azonnal elérhetővé vált a Hugging Face Diffusers könyvtárában. Az új modell egyetlen architektúrában egyesíti a szövegalapú képgenerálást és a több referenciaképet használó képátalakítást. A rendszer szöveges leírással határozza meg a képelemek elrendezését, ahol a felhasználó pontos koordinátákkal adhatja meg a tárgyak vagy a módosítandó területek helyét.
A FLUX.3 szövegfeldolgozásért felelős komponense a Qwen3-VL modellre épül, és az úgynevezett flow-matching elven működik. A lepárolt kiadás 16 lépésben hoz létre 1024×1024 pixeles képet, míg a finomhangolásra szánt alapmodell 50 lépést használ. A modell bf16, fp8r és nvfp4 kvantált formátumokban is letölthető, a betanításhoz pedig beépített DreamBooth LoRA szkriptek állnak rendelkezésre.
Az eszköz nyílt forráskódú súlyai és a kipróbálásra szolgáló felület a Hugging Face platformján érhetők el. A futtatáshoz szükséges kód a legfrissebb `diffusers` és `transformers` Python-csomagok telepítésével használható.
Az eredeti szöveg (Hugging Face)
TL;DR Contents Getting started What FLUX.3 can do Text-to-image Image-to-image Layout control Base model and quantized variants Base model Single-file format checkpoints LoRA finetuning What it costs to run Memory Speed The architecture From pixels to tokens From prompt to tokens Two stages Inside one block Every token gets its own timestep Positions in four dimensions The denoising loop Modular Diffusers See the workflows Encode once, generate many Add a step inside the loop Try it Black Forest Labs FLUX.3 Image Klein is here, and it runs in Diffusers from day one. One model does text-to-image and multi-reference image-to-image, and it adds something new, the layout control. You write bounding boxes into the prompt to say where each part of the image should go, or which parts of a reference image to move.
Both open checkpoints are in the FLUX.3 Image Klein collection:
You can head on to the Hugging Face space and generate some images right now!
Install the latest diffusers and transformers (for the Qwen3-VL text encoder):
Load the distilled checkpoint and generate an image:
FLUX.3 responds best to a full scene description with the subject, the setting, the light, the composition, and the style. The official prompting guide from Black Forest Labs has more tips. The distilled checkpoint generates a 1024 × 1024 image in 16 steps by default. Pass height, width, or num_inference_steps to change them.
Pass a reference image (or more) with the image argument. Write the prompt as a caption of the image you want, with the change included.
With several references, refer to them in the prompt as <ref_image_0>, <ref_image_1>, ... in the order they are passed to image:
Add a JSON list of regions after the caption. Each region has an id, a bbox, and a desc, and the caption mentions it as <id>.
Boxes are written as [y_min, x_min, y_max, x_max], y first, on a 0–1000 scale. Because the scale is normalized, the same box works at any image size.
For edits, each region says where its content comes from (src_bbox in the reference) and where it goes (tgt_bbox in the output). A region with the same two boxes stays where it is.
This example moves the knitted knight up and to the left, and pins the backdrop, the cliff, the needle, the dragon, and the fire:
Everything above uses the distilled checkpoint. The base checkpoint runs 50 steps with classifier-free guidance, and it's the one to fine-tune (see LoRA finetuning).
Black Forest Labs also ships the transformer as a single file, in bf16 and in two quantized formats: fp8r (rowwise FP8) and nvfp4. Load any of them with from_single_file, then load the other components by name so the bf16 transformer isn't downloaded too.
For the unquantized single file, use https://huggingface.co/{model_id}/blob/main/model.safetensors and drop quantization_config.
FLUX.3 works for both text-to-image and image-to-image, which makes it a great base for custom LoRAs. diffusers ships DreamBooth LoRA training scripts for both tasks (you can also train with Ostris' AI Toolkit).
Here is a text-to-image example that learns a tarot-card style from the multimodalart/1920-raider-waite-tarot-public-domain dataset, triggered with trtcrd:
To reduce memory usage, add a few optimization flags:
--gradient_checkpointing recomputes activations in the backward pass (lower VRAM, slightly slower), --cache_latents pre-encodes the images and frees the VAE, and --use_8bit_adam keeps the optimizer states in 8-bit.
A couple of knobs are worth experimenting with. The grid below trains the same tarot style with different LoRA ranks and optimizers; the adapters are on the Hub at diffusers-modular/flux3-tarot-lora. More training features and configurations are documented in the DreamBooth FLUX.3 README.
For instruction-based editing, the image-to-image script (train_dreambooth_lora_flux3_klein_img2img.py) trains on (reference image, target image, instruction) triplets, where the reference is addressed in the prompt as <ref_image
