This Model was trained with around 50 Images. The model was trained with the character having red or blue eyes, so it needs to be specified in the prompt. The model was trained with images that are photorealistic and images that were anime inspired, so depending on the checkpoint the model should be able to do both. V1 isn't trained with nsfw images, so the results are not really good for those images. I reccomend a high cfg scale 5&lt;. When using SDXL the model reaches a high consistency in the character design. When using pony the consistency goes down but nsfw images turn out better (not awesome but better). Update: V2 has some R rated images in the training data set but still no nsfw images. Generally V2 should be more versatile, especially when it comes to different poses and camera angles, than V1.

Lea by Sergeandgreen

It's the first attempt at creating a model for this character, meaning this model might still have a lot of room for improvement. But i think the initial attempts with this model were quite good. I recommend a high cfg scale 5&lt;. The model isn't trained for nsfw, and has problems generating nsfw. This will be fixed in future versions but currently the results are a similar to flux, if you manage to get nsfw images.

Files with SHA256: 5e736713530c1fd61ad56f3f4dffb1c92ee24eeff0799c52508dddd6f872943b