This is a simple crude workflow I did so I could use FLUX on my pc without having to wait 20 minutes (or more if upscaled) for each image generation.So, I decided using the Schnell version, then load an SDXL model to upscale and, if needed, apply face detailer.*If you run into an allocation error when using the face detailer, just close the message and hit "Queue Prompt" again (that's why I like to use fixed seeds).My PC:i5-12400FRTX 3060 12Gb VRAM32Gb RAMHope this helps someone.

FLUX Low RAM/VRAM Workflow: With Upscaler + FaceDetailer

Added a ControlNet group, so I could denoise more without the output going crazy;Added an InPaint group, to remove unwanted objects;No performance loss on my pc.

Files with SHA256: 2d96f858c7ab666db4d273dc21dba5a0f5da7cabe5f6fa1912ad95a2b438ff2a