Type | Workflows |
Stats | 409 0 |
Reviews | (32) |
Published | Sep 4, 2024 |
Base Model | |
Hash | AutoV2 9A7B39852E |
I was asked to share my ComfyUI Workflow - especially for one of my uploaded images. So here we go: this is one of my Workflows I use for FLUX1 Dev. The implemented Upscaling is very basic, you have to keep the denoising by 2.4 or lower. The Promting is a bit more complex and works best with Wildcards - feel free to experiment. I will share mine, when I did a bit of a cleanup so that others might understand what could work well together...
It also works with FLUX1 Schnell and the FP8 Variants. For other Checkpoints like GGUF or NF4 you have to change the Model Loader(s). If you use this Workflow with FLUX1 DEV it will use 94% of 24GB VRAM and it somehow stays there even with 2x2x Upscale.
If the text above doesn't make sense - tell me - I am not a native english speaker and did not use an LLM ;-)