Uploaded September 2026 | Updated November 2026, 1 day ago
Qwen Image 2.1 now has a ControlNet. I tested the new Qwen Image 2.1 Fun ControlNet Union in ComfyUI with pose, depth and inpainting, and I show you what worked and what failed.
One model file gives you 8 control types plus inpainting. When a pose came out wrong and a hand came out twisted, two simple fixes solved it for me.
What I test in this video:
- Pose with DWPose: why a vertical and a horizontal image gave me the wrong pose, and how making both images the same size fixed it
- A hard lunge pose: the twisted hand, and how describing the pose in the prompt and lowering the strength from 1 to 0.8 fixed it
- Depth with Depth Anything V2: a living room changed into a dark room at night, then into an abandoned place, with the furniture in the same position
- A scribble trick (without the ControlNet): a drawing turned into a realistic image, and a woman turned into a zombie
- Inpainting: a top replaced with a T-shirt, and a fruit bowl replaced with a glass vase of white flowers
About the model:
The file is 7.55 GB. Save it in the model_patches folder inside your ComfyUI models folder. It supports Pose, Depth, Canny, HED, Lineart, MLSD, Scribble and Grayscale, plus inpainting.
Important: At the time of recording, the "Apply Qwen Image 2.1 Fun ControlNet" node is not in the normal ComfyUI release. It comes from Kijai's pull request. Without that version or a later update, ComfyUI will show the node as missing.
Resources:
Model: huggingface.co/alibaba-pai/Qwen-Image-2.1-Fun-Controlnet-Union
Kijai's ComfyUI pull request: github.com/Comfy-Org/ComfyUI/pull/16519
My workflow: patreon.com/ComfyuiWorkflowBlog/posts/qwen-image-2-1-170530939
My previous Qwen Image 2.1 video: youtube.com/watch?v=4Xw56Xxe0-k
youtube.com/watch?v=kz8ttx0KAyA
Which control type should I test next: Canny, HED, Lineart, MLSD or Grayscale? Tell me in the comments.
#QwenImage #ComfyUI #ControlNet
Qwen Image 2.1 now has a ControlNet. I tested the new Qwen Image 2.1 Fun ControlNet Union in ComfyUI with pose, depth and inpainting, and I show you what worked and what failed.
One model file gives you 8 control types plus inpainting. When a pose came out wrong and a hand came out twisted, two simple fixes solved it for me.
What I test in this video:
- Pose with DWPose: why a vertical and a horizontal image gave me the wrong pose, and how making both images the same size fixed it
- A hard lunge pose: the twisted hand, and how describing the pose in the prompt and lowering the strength from 1 to 0.8 fixed it
- Depth with Depth Anything V2: a living room changed into a dark room at night, then into an abandoned place, with the furniture in the same position
- A scribble trick (without the ControlNet): a drawing turned into a realistic image, and a woman turned into a zombie
- Inpainting: a top replaced with a T-shirt, and a fruit bowl replaced with a glass vase of white flowers
About the model:
The file is 7.55 GB. Save it in the model_patches folder inside your ComfyUI models folder. It supports Pose, Depth, Canny, HED, Lineart, MLSD, Scribble and Grayscale, plus inpainting.
Important: At the time of recording, the "Apply Qwen Image 2.1 Fun ControlNet" node is not in the normal ComfyUI release. It comes from Kijai's pull request. Without that version or a later update, ComfyUI will show the node as missing.
Resources:
Model: huggingface.co/alibaba-pai/Qwen-Image-2.1-Fun-Controlnet-Union
Kijai's ComfyUI pull request: github.com/Comfy-Org/ComfyUI/pull/16519
My workflow: patreon.com/ComfyuiWorkflowBlog/posts/qwen-image-2-1-170530939
My previous Qwen Image 2.1 video: youtube.com/watch?v=4Xw56Xxe0-k
youtube.com/watch?v=kz8ttx0KAyA
Which control type should I test next: Canny, HED, Lineart, MLSD or Grayscale? Tell me in the comments.
#QwenImage #ComfyUI #ControlNet










