FLUX is an open-weight, guidance-distilled model developed by Black Forest Labs.

Minimum Hardware Requirements: 24GB VRAM, 32GB RAM

Steps:

1- Update your Comfy UI

2- Download t5xxl_fp8_e4m3fn.safetensors and clip_l.safetensors in your ComfyUI/models/clip/ directory:  https://huggingface.co/comfyanonymous/flux_text_encoders/tree/main

If you have more than 32GB ram, FP16 version is recommended and you can use t5xxl_fp16.safetensors instead. For lower memory usage but the t5xxl_fp8_e4m3fn is recommended.

3- Download VAE and put it under the ComfyUI/models/vae/ folder:  https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.sft

4- Download the flux1-schnell.sft and place it under ComfyUI/models/unet folder:   https://huggingface.co/black-forest-labs/FLUX.1-schnell/tree/main

You can use the flux-dev version as well but schnell versions offers 1 to 4 steps and it has Apache 2.0 licence.

Huge thanks to Black Forest Labs and Comfy UI teams!

For further details please check:  

https://comfyanonymous.github.io/ComfyUI_examples/flux/

https://blackforestlabs.ai/