!!! attention 
workflow is not focused on speed!!!
only on quality!
works on 4070ti super 16 gb -- batch=4  4.24s/it ,1024x1024

v2 add text combined and switch promt (llama - llama+text - text)
This workflow allows you to transfer style at will, generate hints at will.
 Stop generation with preview and composition selection, face detailer and upscaler,
  all stages with preview and selection option, default delay 300 ms,
   if you do not select send to the next stage everything, 
by default batch size-4, style weight is adjustable, 0.2-0.4

works great with character lora

manager-install missing custom nodes

STYLE

-------------------------------------------

flux1-redux-dev

ComfyUI-Stable\ComfyUI\models\style_models

------------------------------------------

sigclip_vision_patch14_384_F1

ComfyUI-Stable\ComfyUI\models\clip_vision

------------------------------------------

PROMT

install

https://ollama.com/download

command com - ollama run llava

------------------------------------------

t5xxl_fp16

clip_l

ComfyUI-Stable\ComfyUI\models\clip

------------------------------------------

flux1-dev-fp8

ComfyUI-Stable\ComfyUI\models\unet

------------------------------------------