Introduction
This model is fine-tuned from Anima Base 1.0.
It focuses on cute and diverse anime-style outputs while improving overall stability.
Recommended Settings
Steps: 25–30
shift: 4-6
CFG scale: 4–5
Sampler: ER SDE / Euler A / DPM++2M
Notes
If you encounter any issues, please leave feedback so I can try to improve them in future versions.
Description
For v2, DPO training was conducted on approximately 3,000 preference pairs to improve model quality.
FAQ
Comments (4)
pretty good versatile model. i really like it.
Very good and cute model, i watch this for future
This model performs great in all aspects and has a lovely style, but there are a couple of issues: its text generation capability has weakened, and it often fails to produce correct characters. Also, it still has minor issues with hands and feet. Everything else is excellent, and we hope the next version will be even better.
This is an excellent model. Other Anima checkpoints often look impressive on the surface, but in practice, they either do not have much of an impact or do not work well with LoRAs. This checkpoint, however, has a clear personality and works harmoniously. Thank you for your work.