Instructions to use dx8152/Flux2-Klein-9B-Consistency with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use dx8152/Flux2-Klein-9B-Consistency with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline from diffusers.utils import load_image # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("black-forest-labs/FLUX.2-klein-9B", torch_dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("dx8152/Flux2-Klein-9B-Consistency") prompt = "Turn this cat into a dog" input_image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cat.png") image = pipe(image=input_image, prompt=prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
for skincare products writings
Klein itself doesn't have the ability to recognize text, so during editing, it encodes text as an image and then decodes it. Therefore, it's very difficult for text to be identical; perhaps we'll have to wait for a better model.
Thank you
@JzoughI was curious, and surprisingly, using Ideogram JSON prompt builder with any large LLM and piping the result into Klein 9b with reference image and higher resolution produces decent text results. I tried using qwen 3 4b/8b, but it struggles with long prompt, so im using Gemini 3 flash / Gemini 3.5 flash to build intiial prompt and then adjust manually with KJ prompt builder node
The answer for the text clarity is detailed prompt + resolution. This wf uses crop&stich node to give model better resolution to "see" and resolve text label


