Qwen Image 2.1 + ControlNet Union 9-in-1 workflow: 8 CN control modes plus Mask-based local repaint in a single graph. One input image, nine parallel branches, all outputs in one pass.
Main features:
- 9 parallel branches on Qwen Image 2.1 (bf16) with the Fun ControlNet Union patch and the Turbo LoRA (rank 178)
- 7 AIO preprocessor modes at 512: Canny, MiDaS depth, HED, LineArt, M-LSD (structure), DW depth, Scribble
- Structural-guidance branch (raw image condition) and a Mask-based inpainting branch (ImageToMask -> ImageCompositeMasked) for local repaint
- Gemini 3.1 Flash Lite (RHLLMChatNode) auto prompt analysis: describe the source image automatically for consistent prompts across branches (auto prompt reconstruction / image wash)
- Per-branch ManualSigmas, image resize and size-math (pysssss) for matched output dimensions
Suggested workflow:
1. Load the source image; optionally let Gemini reconstruct the prompt automatically
2. Run the graph; all 9 branches (7 control modes + structure + mask inpaint) render together
3. Pick the control mode per task: pose / depth / line / structure for control, mask branch for local edits
Dependencies:
- qwen_image_2.1_bf16.safetensors (base UNET)
- qwen3vl_8b_fp8_scaled.safetensors (CLIP)
- qwen_image_2.1_vae_bf16.safetensors (VAE)
- Qwen-Image-2.1-Fun-Controlnet-Union-ComfyUI.safetensors (Union patch)
- qwen_image_2.1_turbo_lora_avg_rank_178_bf16.safetensors (Turbo LoRA)
- pysssss MathExpression, easy showAnything, JWString nodes, AIO Preprocessor, RHLLMChatNode
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2109549130926755842?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
Description
Qwen Image 2.1 + ControlNet Union 9-in-1 workflow: 8 CN control modes plus Mask-based local repaint in a single graph. One input image, nine parallel branches, all outputs in one pass.
Main features:
- 9 parallel branches on Qwen Image 2.1 (bf16) with the Fun ControlNet Union patch and the Turbo LoRA (rank 178)
- 7 AIO preprocessor modes at 512: Canny, MiDaS depth, HED, LineArt, M-LSD (structure), DW depth, Scribble
- Structural-guidance branch (raw image condition) and a Mask-based inpainting branch (ImageToMask -> ImageCompositeMasked) for local repaint
- Gemini 3.1 Flash Lite (RHLLMChatNode) auto prompt analysis: describe the source image automatically for consistent prompts across branches (auto prompt reconstruction / image wash)
- Per-branch ManualSigmas, image resize and size-math (pysssss) for matched output dimensions
Suggested workflow:
1. Load the source image; optionally let Gemini reconstruct the prompt automatically
2. Run the graph; all 9 branches (7 control modes + structure + mask inpaint) render together
3. Pick the control mode per task: pose / depth / line / structure for control, mask branch for local edits
Dependencies:
- qwen_image_2.1_bf16.safetensors (base UNET)
- qwen3vl_8b_fp8_scaled.safetensors (CLIP)
- qwen_image_2.1_vae_bf16.safetensors (VAE)
- Qwen-Image-2.1-Fun-Controlnet-Union-ComfyUI.safetensors (Union patch)
- qwen_image_2.1_turbo_lora_avg_rank_178_bf16.safetensors (Turbo LoRA)
- pysssss MathExpression, easy showAnything, JWString nodes, AIO Preprocessor, RHLLMChatNode
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2109549130926755842?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