Flux speed workflow+loras+hires fix on 8GB VRAM_V4
created 8 months ago
upscale
lora
47 nodes

633

121

Outputs
example image output optimized_comfy_workflows_user_uploads/e5b63680-9264-4c2e-be0b-a166e08970c5/assets/ComfyUI_temp_jqqrp_00007_.webp for workflow.
Description

A scheme for generating images in 3 minutes including loras and upscaling.

Three minutes rendering on an RTX-3060 12GB graphics card.

I recommend 8+4+4 steps.

The first render will be long, don't worry, the models will load into memory.

On good video cards, it will be counted instantly!

Instructions for running inside.

Thank you all.

Мore examples:

https://t.me/artgptchat

Built-in nodes
SamplerCustomAdvanced
BasicGuider
RandomNoise
CLIPTextEncodeFlux
Custom nodes
DualCLIPLoader
UNETLoader
VAEDecode
easy clearCacheAll
BasicScheduler
ImageUpscaleWithModel
ImageScaleBy
UpscaleModelLoader
VAELoader
KSamplerSelect
SaveImage
SDXLEmptyLatentSizePicker+
VAEEncode
0
0
0
0