Whenever you want to do some kind of die hard editing to images, you of course need top level blending skills and patience to get satisfied results. ByteDance has recently introduced Uno (Unity and Novel Output) to solve this problem.
An innovative AI framework designed for in-context image generation that leverages FluxDev-based model to creatively merge multiple image references into cohesive outputs. The model is registered under Flux Dev license.
The model can handle maximum 4 images in one go as the input reference and produces an output. You can go in depth by accessing their research paper.
This development represents a significant step forward in AI-driven image synthesis, particularly for applications requiring the blending of styles, objects, and virtual try-ons, such as e-commerce and digital content creation.
Installation
1. Install ComfyUI if are new to it.
2. Update it from the Manager section by selecting "Update ComfyUI" if you are an old user.
3. Download the UNO Flux LoRA model (dit_lora.safetensors) from Bytedance hugging face repository. Create the UNO folder inside "comfyui/models" directory. Save this into your "comfyui/models/UNO/" directory.
4. Clone the repository for UNO model using the following command into your command prompt:
git clone https://github.com/HM-RunningHub/ComfyUI_RH_UNO.git
5. Next, you also need the Flux official model by Black Forest Labs, Vae, Clip and T5 models downloaded that should be saved into their basic respective folder. If you haven't done the basic setup, follow our Flux installation tutorial.
Folder structure:
ComfyUI/models
flux/
FLUX.1-schnell/
text_encoder/
tokenizer/
text_encoder_2/
tokenizer_2/
unet/
flux1-schnell.sft
flux1-dev.sft
vae/
ae.safetensors
UNO/
dit_lora.safetensors
6. Restart your ComfyUI to take effect.
Workflow
1. The respective workflow can be found inside your "ComfyUI/custom_nodes/ComfyUI_RH_UNO/examples" folder as flux-uno.json file.
2. Drag and drop into ComfyUI.
Upload your reference images. You can put maximum four images.
Put your relative prompt.
Prompt used: a beautiful woman with sunglasses, wearing red dress in newyork city street
Steps: 25
CFG: 4
 |
| Reference 1 |
 |
| Reference 2 |
 |
| Reference 3 |
 |
| Reference 4 |
 |
| First Output generation |
 |
| Second output Generation |
