This is my first time training a LoRA for Wan2.2. There might be many issues, but it can still produce some pretty good results. (I personally think the results look quite good QWQ)

In the demonstration, this model was loaded for both the high-noise and low-noise stages in ComfyUI, with a high-noise strength of 1.0 and a low-noise strength of 1.5. (I'm not sure if this is the absolute best setup, but it worked pretty well for me.)

Based on my testing, both 16:9 and 9:16 aspect ratios yield good results.

(The prompts shown might have slight discrepancies since I generated too many images and might have misremembered. While Wan2.2 supports English, it's best to use Chinese for the trigger words.)

The workflow I used is linked below ↓

https://civitai.com/articles/18232/wan22lightning-workflow

This model might not be perfect yet, I'll try to update it again when I have time :)

Thank you for using it, and have fun! :)

Reprinted from: Wan2.2/Minimalist Solid Color Anime Style