Uploaded September 2026 | Updated November 2026, 4 days ago
Qwen Image 2.1 is a new open-weight image model, and ComfyUI supports it natively. In my test, it looked more realistic than ChatGPT, but it also made mistakes with hands, feet, and a cup. I show you the one simple setting that fixed most of them.
Then I test the new features: transparent cutouts, character sheets, a storyboard, camera angles, and an outfit swap using three images.
What I cover:
- The files you need: INT8 (about 7.26 GB) or BF16 (about 14.2 GB), the text encoder (about 9.35 GB), and the Qwen Image 2.1 VAE (about 676 MB)
- A simple workflow with the new Text Encode Qwen Image 2.1 node
- How I fixed the hands and feet in my test
- Why I use 40 steps instead of the default 25, and what changes at CFG 2
- A 7-part prompt structure, and why the "no" words go in your main prompt at CFG 1
- Two prompt helpers: Gemma 4 and Qwen's official prompt enhancer (Qwen Image 2.1 PE)
- Text and infographics
- Character sheets and a 4-frame storyboard
- Background removal, and cutting out a single object with transparency (RGBA)
- Image editing prompts: new background, object removal, clothes swap, and a new camera angle
- 8 camera angles of the same scene
- Where ChatGPT is still better (small details and text), and how Gemini did
My test setup: RTX 5090, BF16 model, Qwen3-VL 8B text encoder, 25 steps, CFG 1, Euler sampler (the official ComfyUI starting point). The first image took around 10 seconds.
License: Qwen Image 2.1 uses the Qwen Research License. It is for research and non-commercial use. If you want to use it for client or commercial work, read the license first.
Resources
Model files for ComfyUI (Hugging Face): huggingface.co/Comfy-Org/Qwen-Image-2.1
Official ComfyUI guide and workflows: docs.comfy.org/tutorials/image/qwen/qwen-image-2-1
Qwen Image 2.1 on GitHub: github.com/QwenLM/Qwen-Image-2.1
License: huggingface.co/Qwen/Qwen-Image-2.1/blob/main/LICENSE
My workflow: patreon.com/ComfyuiWorkflowBlog/posts/qwen-image-2-1-3-170249435
I will make more videos about this model. Subscribe so you do not miss the next one.
#ComfyUI #QwenImage #AIImageGeneration
Qwen Image 2.1 is a new open-weight image model, and ComfyUI supports it natively. In my test, it looked more realistic than ChatGPT, but it also made mistakes with hands, feet, and a cup. I show you the one simple setting that fixed most of them.
Then I test the new features: transparent cutouts, character sheets, a storyboard, camera angles, and an outfit swap using three images.
What I cover:
- The files you need: INT8 (about 7.26 GB) or BF16 (about 14.2 GB), the text encoder (about 9.35 GB), and the Qwen Image 2.1 VAE (about 676 MB)
- A simple workflow with the new Text Encode Qwen Image 2.1 node
- How I fixed the hands and feet in my test
- Why I use 40 steps instead of the default 25, and what changes at CFG 2
- A 7-part prompt structure, and why the "no" words go in your main prompt at CFG 1
- Two prompt helpers: Gemma 4 and Qwen's official prompt enhancer (Qwen Image 2.1 PE)
- Text and infographics
- Character sheets and a 4-frame storyboard
- Background removal, and cutting out a single object with transparency (RGBA)
- Image editing prompts: new background, object removal, clothes swap, and a new camera angle
- 8 camera angles of the same scene
- Where ChatGPT is still better (small details and text), and how Gemini did
My test setup: RTX 5090, BF16 model, Qwen3-VL 8B text encoder, 25 steps, CFG 1, Euler sampler (the official ComfyUI starting point). The first image took around 10 seconds.
License: Qwen Image 2.1 uses the Qwen Research License. It is for research and non-commercial use. If you want to use it for client or commercial work, read the license first.
Resources
Model files for ComfyUI (Hugging Face): huggingface.co/Comfy-Org/Qwen-Image-2.1
Official ComfyUI guide and workflows: docs.comfy.org/tutorials/image/qwen/qwen-image-2-1
Qwen Image 2.1 on GitHub: github.com/QwenLM/Qwen-Image-2.1
License: huggingface.co/Qwen/Qwen-Image-2.1/blob/main/LICENSE
My workflow: patreon.com/ComfyuiWorkflowBlog/posts/qwen-image-2-1-3-170249435
I will make more videos about this model. Subscribe so you do not miss the next one.
#ComfyUI #QwenImage #AIImageGeneration










