YESSSS!

it is possible to run FLUX.1 on an RTX 3060 12 GB VRAM and 32 GB of system RAM.

1 - Download the 11.9 GB model from:

https://huggingface.co/Kijai/flux-fp8/tree/main

move it to comfyUI/models/unet

I use DEV with 20 steps but we can run Schnell at 4 steps

Note: Schnell is more speed and less quality but upscalers can fix that.

2 - Download the fp8 clip and clip_l from:

https://huggingface.co/comfyanonymous/flux_text_encoders/tree/main

move them comfyUI/models/clip

3 - Download the vae model from:

https://huggingface.co/black-forest-labs/FLUX.1-dev/tree/main/vae

rename the file to flux_vae.safetensors an move to comfyUI/models/vae

all done! -

Launch comfyUI with argument --lowvram after main.py in the terminal or add it to your .bat file

Load the workflow and close any unnecessary browsers and windows.