Trained with 18 images, using just the standalone keyword is sometimes enough, but using more specific descriptions yields better results.
syoka \(vtuber\), white dress, halo,
You can also generate her mascots by just writing "sheep mascots".
Initial release
syokaanima2.safetensors