Mirevyn v3
Anime art generation LoRA for MOSS-Image
1 min readupdated
This model is a LoRA (a small add-on for a base model) trained using the Ostris AI Toolkit on top of OpenMOSS-Team/MOSS-Image-Base. The checkpoint is designed to generate detailed illustrations in a recognizable anime style with a focus on quality and composition. To trigger the effect, use a specific trigger word at the beginning of the prompt. This model is useful for artists and enthusiasts creating stylized images in ComfyUI or AI Toolkit.
Key features
- Anime style built on MOSS-Image Base with a LoRA rank of 32/32
- Signature trigger token
mirevynto control styling - Trained for 2,500 steps using the AdamW 8-bit optimizer
- Precise tuning of generation parameters for high detail
What you can do
- Generate stylized anime illustrations and characters in ComfyUI
- Use the LoRA adapter with the Ostris AI Toolkit for local generation
- Experiment with adapter strength and different seeds to get the desired result
Example
Sample prompt from the model card: mirevyn, an adult cyclist in a yellow raincoat crosses a wet city street at night.
Read more
DetailsBase model: OpenMOSS-Team/MOSS-Image-Base
Training tool: Ostris AI Toolkit (2,500 steps)
Recommended settings: CFG 4.5, 45 steps, Euler sampler, LoRA strength 1.0
Training resolution: 1024 (area-based aspect-ratio buckets)










Comments