Third Diffusion (Style LoRA)

Third Diffusion

The style mostly applies to subjects.

Please share some images. ♥️

Checkpoints

• Version 2: Cartunafied 3 (Recommended)

• Version 1: NAI

Training Description (Latest Version)

• Images: 250

• Resolution: 768 / 768

• Average Tags Per Image: 60

• Caption Style: Comma Separated Booru Tags Without Underscores

• Steps: 7500

• Rank: 128

• Alpha: 32

• Clip Skip: 2

• Tokens: 225

• Precision: FP16

• Optimizer: Adafactor

• Learning Rate: 0.0001

• Scheduler: Cosine With Restarts

• Noise Offset: 0.1 (Original)

• Min SNR Gamma: 5

• Shuffle Captions: True

• XFormers: True

Note

The training images are from the game Third Crisis by Anduo Games (NSFW).

Feel free to check it out!

Donations

Ko-Fi