ComfyUI: Creating a character scene with DW Pose and Qwen2-VL

Step-by-step

1. Load the reference image via URL, local file path, or the ComfyUI input folder
2. Connect the image to an Image Switch node to toggle between references
3. Run the image through the Qwen2-VL node to autogenerate a text description
4. Disable the Quant Image Edit model group to extract the caption first
5. Re-enable the Quant Image Edit group using the latest 200511 model
6. Input the DW pose map into the second image slot
7. Enter the desired visual style and action scene details into the text box

Options

• Load images from a URL
• Load images from a local file path
• Load images from the ComfyUI input folder
• Toggle between different references using an Image Switch node

Watch out for

• The Quant Image Edit model group must be disabled initially if you only want the caption without running a full edit

Tips

• Copy the autogenerated description from Qwen2-VL to use as parts of your final prompts later
• Use an Image Switch node for convenience when working with multiple reference images

Highlights

“This is super helpful because I can copy that description and use parts of it in my final prompts later.”