Too lazy to write the prompt, use an image instead of a prompt.
The vision is brought by Qwen 3.5 with MiniMax H3 Image.
Performance 2048x2048: About 16 seconds 5070ti, 32 sec 5060ti 16g (2Nd gen/warm).
If the input image is below 4 MP, the image is upscaled to 4 MP because MiniMax H3 gives better results at 4 MP.
No custom nodes, no custom code, 100% native, so ComfyUI won't crash when you do an update. All necessary files are indicated in the workflow.