Version 2 of SUISUI from Wuthering waves

✦ Improvements

• Improved the dataset, the lora was trained on 43 New images from different artist with NSFW images as well, therefore giving it a much better NSFW generations, especially the clothing, in V1 the neck and some other clothing was generated even if you drop the weights and not include them in the prompt, that has been fixed

• More detailed, the Eye color is fixed, she has the beautiful blue and yellow eyes, and her forehead tacet mark has also been fixed

• In general this Lora has more detail and does not affect the art style and is more consistent

• Better NSFW generation

• the original outfit is more detailed now, even the back side, you can use prompts like "backless outfit" to get the detailed back side look

✦ Recommended Settings

• Trigger Word: Suisui — use alongside descriptive tags for best results

• Weight: Start at 0.6–1.0 and adjust to taste. Lower weights (~0.6) blend her look more naturally; higher weights (~1.0–1.2) push stronger likeness

• Sampling Steps: 12 (Euler a) (note that I am using the anima-turbo-lora-v0.2 so i am using lower steps and 1 CFG)

• CFG Scale: 1

• Resolution: 1216x1536 (tested)

• For the alternate clothing please remove the clothing prompts from the default trigger words and use "alternate costume/ alternate outfit" and mess with the weights of other clothes prompt (eg: shirt:1.2, pants:1.3), and in general try to use different weights and find a good sweet spot for your generations as I have not tested this with multiple loras

✦ Known Limitations

• Might be inconsistent with alternate outfit/ clothes at a higher weight

✦ Feedback & Suggestions Welcome!

This is my LoRA and I'm always looking to improve! If you try this LoRA, I'd love to hear from you:

• Share your generations — post images in the reviews/comments so others (and I) can see what's possible with different prompts and styles!

• Report issues — if something looks off (e.g., color bleeding, artifacting, bad anatomy at certain weights), let me know along with the settings you used so I can investigate

• Suggest improvements — have ideas for better trigger words, tag combinations, or training updates? Drop a comment, I read them all!

• Rate and review — your ratings help others discover this LoRA and help me understand what's working and what isn't

Your feedback will will help a lot. Thank you for trying it out!