Sign In

Simple Workflows for Qwen-Ultra-Realistic-NSFW (BF16 40GB) and Qwen model

5

This is a simple workflow for running the Copax Timeless model (Qwen-Ultra-Realistic-NSFW (BF16 40GB) versionl) and the Qwen model.

The BF16 40GB version can run on an RTX 3090 with a 2K image resolution. I haven’t tested it yet with larger image sizes or on lower-end GPUs.

image.png

During execution, it consumes a very large amount of VRAM and also relies heavily on system RAM, so machines with limited RAM may not be able to run it properly—please keep this in mind.

I tested the demo at a 2K image size. If your system has stronger hardware, it should be possible to run at higher image resolutions.

workflow (1).png

ComfyUI/models/Diffusion_models ( BF16 40GB)

https://civitai.com/models/118111?modelVersionId=2538295

ComfyUI/models/vae (250Mb)

https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors?download=true

ComfyUI/models/clip (9GB)

https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b_fp8_scaled.safetensors?download=true

5