(Default) Qwan Image workflow extended with:
LLM Prompt (Ollama DeepSeek R1)
LoRA
FaceDetailer (SDXL Turbo - default model)
Upscaling options:
(pixel) SD1.5 tile upscale with ControlNet
(pixel) SDXL tile upscale with ControlNet
(pixel) SDXL Lightning tile upscale with ControlNet
(pixel) Flux-Dev (Q8_0)
(latent) WAN2.2-T2V-LowNoise (Q4_K_M)
Additional notes from testing LoRA's, settings, etc.
Default settings require 24GB of VRAM.