Anireal is an anime-realism style LoRA trained for MOSS Image. It is designed to preserve recognizable anime character identity while adding more natural skin, detailed hair, textured fabric, cinematic lighting, and greater depth to environments.
This is my first public LoRA release. It is not perfect yet: hands, very small details, and text in complex scenes may still be inconsistent. I plan to keep improving the dataset and training settings in future versions. Feedback and example images are very welcome.
No trigger word is required.
Recommended settings:
- LoRA strength: 0.8-1.0 (showcase images use 1.0)
- Steps: 50
- CFG Scale: 3
- Sampler: Euler
- Scheduler: sgm_uniform
- Resolution: 1024 x 1536
All showcase images include ComfyUI generation metadata.
Description
Initial public release. This is my first LoRA for MOSS Image. The model is still being improved, and future versions will focus on better consistency, details, and prompt response.
FAQ
Comments (9)
我喜欢你♿
么么么么
很清新啊
老大老大
Wow it works great! 🥰
Thank you for your comment! This is my first time training a LoRA. I’m really optimistic about MOSS—I’ve experimented with both Illustrious and Anima, but I feel that MOSS has a very promising future!❤️
NB
Excellent. I look forward to a future version with further refinements to details like the hands.
Looks cool I’ll try and drop a prompt!