Verified: a year ago
SafeTensor
Type | |
Stats | 5,847 82,338 |
Reviews | |
Published | May 1, 2023 |
Base Model | |
Training | Steps: 3,000 Epochs: 10 |
Trigger Words | vtuber-halfbody vtuber-fullbody |
Hash | AutoV2 427C1C87DD |
LoRa created with 100 images of full body (with tw vtuber-fullbody) and upper body vtuber models (with tw vtuber-halfbody).
Add full body or upper body at the end of your promt if you have problem with the result.
White background or simple background can sometime work but it tend to not change anything. If you want to have the character without the background, I suggest you to use an AI crop tool like Segment Anything from Meta AI.
To have better quality images, use hires fix. and upscalers (especially R-ESRGAN 4x+ Anime6B)
If you manage to have some great results or if you want to suggest some improvement for the model, you can send me feedback in the comment.