ComfyUI Workflow Qwen Image
Any feedback would be appreciated.
📂 Required Models
Diffusion Model: qwen-image-2512-Q8_0 .../ComfyUI/models/diffusion_models/
Text Encoder: qwen_2.5_vl_7b_fp8_scaled .../ComfyUI/models/text_encoders/
VAE: qwen_image_vae .../ComfyUI/models/vae/
📂 Optional Models
Turbo LoRA: Wuli-Qwen-Image-2512-Turbo-LoRA-4steps-V1.0-bf16_ComfyUi .../ComfyUI/models/loras
Upscale Model: 4x_foolhardy_Remacri.pth .../ComfyUI/models/upscale_models
Detection Model: face_yolov9c.pt .../ComfyUI/models/ultralytics/bbox
SAM Model: sam_vit_b_01ec64.pth .../ComfyUI/models/sams
🧩 Required Custom Nodes
🟩 ComfyUI-Manager (by Comfy-Org)
https://github.com/Comfy-Org/ComfyUI-Manager🟩 rgthree-comfy (by rgthree)
https://github.com/rgthree/rgthree-comfy🟩 ComfyUI-Easy-Use (by yolain)
https://github.com/yolain/ComfyUI-Easy-Use🟩 ComfyUI-Image-Saver (by alexopus)
https://github.com/alexopus/ComfyUI-Image-Saver🟩 ComfyUI-Impact-Pack (by ltdrdata)
https://github.com/ltdrdata/ComfyUI-Impact-Pack🟩 ComfyUI-Impact-Subpack (by ltdrdata)
https://github.com/ltdrdata/ComfyUI-Impact-Subpack
⚙️ Recommended Settings
Steps: 4 (with LoRA).
CFG: 1.0 (with LoRA).
Sampler: Euler
Scheduler: Simple
"1:1": (1328, 1328),
"16:9": (1664, 928),
"9:16": (928, 1664),
"4:3": (1472, 1104),
"3:4": (1104, 1472),
"3:2": (1584, 1056),
"2:3": (1056, 1584),Experiment and enjoy!
Description
Upload of V1
(This is not an exhaustive changelog)
Added V1:
Wildcard support
Upscale
AuraFlow Model Shift
Image Saver with Metadata
LoRA Loader
Detailer
Multipass
FAQ
Comments (4)
The upscale function and the 2nd pass are probably not both needed. If you want to do a 2nd pass, it upscales by 1.5, so upscaling on top of that.. you might end up with a 50 MB .png file. Also, I didn't use the GGUF and found when I was doing the face detailer pass my machine could only load a gig or so, the rest of the model was offloaded, so my process slowed to a crawl. Not sure why exactly, but if I disabled the detailer, I still got a very detailed face in the end. Thanks for the workflow!
Yeah, using both the upscale and multipass together is overkill. It's intended to be more of a choice between one or the other. But if someone really wants to use both, I would suggest decreasing the latent upscale value.
@Legendaer xD yeah, i liked the look of both on, and ran a 20 prompt batch at once. firefox started lagging a bit from displaying the results in assets
Brilliant WF!