V6 Release🎉
I originally planned to end development with V5, but after more testing and feedback I ended up improving a few things enough to justify one last update.
So here’s V6.
No trigger word is required.
Recommended strength: 0.7 – 1.0
Hi-Res Fix and ADetailer are recommended.
Here are a few helpful negative prompts for my LoRA:
(white theme:1.2), bokeh, (blurry:1.2)Description
V3 Summary
Improved style consistency and overall quality.
Cleaner linework, better colors, and more stable results.
Refined dataset for fewer errors.
FAQ
Comments (5)
I didn't see any difference. Maybe I'm doing something wrong?
This version is based on the same training material, with a few images replaced and slight quality improvements.
The adjustments were made to better align the output with the original dataset style and achieve more consistent results.
@Buccha I mean, with or without her, there's no difference.
@panarinmoskow116 If you’re using WAI-Illustrious, you probably won’t notice much difference — except for things like cleaner lineart, better highlights, and slightly stronger contrast.
The bigger change is in anatomy, especially around penis details, since the LoRA was trained on that.
If you use a different checkpoint, you’ll definitely see a much clearer difference.
I uploaded an earlier V3 version yesterday (04/01/2026), but it had several noticeable issues.
These have now been largely fixed. Minor artifacts like extra fingers or toes may still occur occasionally — I’ve reduced them as much as possible for now.