Qwen Image 2.1 + ControlNet Union depth control parameter matrix (75-set showdown). One reference image drives a full strength / start / end grid of depth-conditioned generations, so you can pick the control level that keeps the subject while reshaping the structure.
Main features:
- Single reference image; DepthAnything V2 (ViT-B, 1024) generates the depth map
- ControlNet Union patch in Depth mode: 75 branches = strength {0.2, 0.4, 0.6, 0.8, 1.0} x start {0, 0.25, 0.5} x end {0.6, 0.7, 0.8, 0.9, 1.0}
- Every branch: Qwen Image 2.1 UNET, 25 steps, CFG 3.5, euler/simple, fixed seed for direct comparison
- 78 outputs decoded and previewed in one pass (75 matrix outputs + references)
Suggested workflow:
1. Load your reference image and edit the prompt
2. Run the graph; all 75 depth-control variants render with the same seed
3. Compare the grid, then copy the winning strength / start / end into your production workflow
Dependencies:
- Qwen-Image_2.1.safetensors (base UNET)
- qwen3vl_8b_int8_convrot.safetensors (CLIP)
- qwen_image_2.1_vae_bf16.safetensors (VAE)
- Qwen-Image-2.1-Fun-Controlnet-Union-ComfyUI.safetensors (Union patch)
- depth_anything_v2_vitb.pth + DepthAnything V2 preprocessor node
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2109551145463549954?inviteCode=rh-v1111
Bilibili tutorial video: https://www.bilibili.com/video/BV14zp96REo9/
RunningHub registration with invite code rh-v1111 (1000 RH coins): https://www.runninghub.ai/?inviteCode=rh-v1111