For my enjoyers of the cute & funny.
This is a style LoRA, use it with character LoRAs!
I look forward to your posts, reviews and comments!
I trained this on frame grabs of the first episode of the anime along with official art.
Ratio of frame grabs to official art, 80:20 (585 Images)
I did not train this lora on character tags, so prompting characters out of this is purely gacha.
I highly recommend you use character LoRAs to prompt you favorites.
Prompt: score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, source_anime,
Negative Prompt: source_furry, source_pony, censored,
You can refer to the provided images for sample prompts.
I recommend using this VAE!
Moved to the 'About this version' tab!
📑 Training Notes
Base Model: Pony Diffusion V6 XL
Training Resolution: 1024
Training Batch Size: 4
Type: LoCon
DIM / Alpha: 16 / 8
Conv Dim / Alpha: 24 / 12
Train On: Unet
Optimizer: AdamW8bit
Scheduler: Rex
Min SNR Gamma: 5
Bucket Resolution: 512 - 2048
Bucket Steps: 64
Noise Offset: Normal (0.025)
Image Count: 585
Steps: 1400
1. 转载模型仅供学习与交流分享，其版权及最终解释权归原作者。
2. 模型原作者如需认领模型，请通过官方渠道联系海艺AI工作人员进行认证。我们致力于保护每一位创作者的权益。 点击去认领
