YourAnimeHaven
Practicing making LoRa
Works pretty well with aDetailers especially face.
v0.1 was used to generate training data for Beta v0.2 and I personally find v0.2 to work better but both are still difficult to prompt and require messing around with cfg and LoRa weights depending on what you're using it for.
Check descriptions on each version for notes.
Popularity
Info
Version Beta v0.2: 1 File
About this version: Beta v0.2
Recommended Settings
LoRa model & clip: 0.65
Euler_a normal
28 steps
2.5 - 5 cfg
Trained on images generated by v0.1
Slight bias towards nudity and difficulty maintaining jewelry count and nail polish while posing but still working very well. Face aDetailer is recommended for best results.
Considering merging Beta v0.2 with v0.1 if I still struggle to get results I'm looking for after playing around with it a bit more otherwise happy with where it's at here for now.
Be careful not to use too many trigger words and start from hair then eyes, then the rest as you need after posing to make sure it doesn't cause distortions, like when trying to get black nails.
Jacket trigger covers most clothing but pants will bring legs and boots in.
3 Versions
Go ahead and upload yours!
