FLUX SUPER Workflow

Flux SUPER Workflow

This workflow is made for Flux and can do:

Text2Img (with or without autoprompt based on the input image)

Img2img (Flux Redux, Controlnet Union, Regular Img2img, everything accesible and easy tweakable with the main knobs)

Cnet Upscale (with noise injection / colormatch)

ExtraFace Detailer (with noise injection / mix / colormatch)

PostProcess (Eveything you need)

• Nice Console Infos

• Minimal clutter, Maximal features

• Bypass groups

• Caching Optimisations, Teacache for faster inference

You can now use FLUX KREA as base model

To enhance the image only, activate img2img with "img2img denoise" to 0. This will bypass the first inference to only use the upscaler and face detailer.

This method is way faster than something like SUPIR and give better details.

Use the "Control_end" as a strength parameter, higher value for lower strenght.

Use the noise "injection strenght" to add more details

Flux dev (scaled)

https://huggingface.co/comfyanonymous/flux_dev_scaled_fp8_test/blob/main/flux_dev_fp8_scaled_diffusion_model.safetensors

T5XXL

https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors

clip l

https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors

vae

https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/vae/diffusion_pytorch_model.safetensors

Flux Redux dev

https://huggingface.co/black-forest-labs/FLUX.1-Redux-dev/blob/main/flux1-redux-dev.safetensors

sigclip_vision_patch14_384.safetensors

https://huggingface.co/Comfy-Org/sigclip_vision_384/blob/main/sigclip_vision_patch14_384.safetensors

FLUX.1-dev-ControlNet-Union-Pro

https://huggingface.co/Shakker-Labs/FLUX.1-dev-ControlNet-Union-Pro/blob/main/diffusion_pytorch_model.safetensors

Flux.1-dev-Controlnet-Upscaler

https://huggingface.co/jasperai/Flux.1-dev-Controlnet-Upscaler/blob/main/diffusion_pytorch_model.safetensors

yolov8m-face.pt

https://github.com/akanametov/yolo-face?tab=readme-ov-file#models

face_yolov8n-seg2_60.pt

https://huggingface.co/jags/yolov8_model_segmentation-set/blob/main/face_yolov8n-seg2_60.pt

Yolo8 face segmentation models are in pickle serialization format and autoflagged as "unsafe" because this is not a secure format, just like ".ckpt"

Be sure to download your models from trusted sources by looking at the profiles and previous commits.

• https://buymeacoffee.com/designedbycrt