Verified: 2 years ago
SafeTensor
Type | |
Stats | 804 |
Reviews | (149) |
Published | Mar 21, 2023 |
Base Model | |
Training | Steps: 100 Epochs: 1 |
Trigger Words | hok style hok-100-128dim-64na-768x.safetensors |
Hash | AutoV2 0CFD4C1BF7 |
This LoRA was trained on GuoFeng3_v32Light as base model using 500 Honor of Kings character artworks (with their logo) so that mean a lot of the generated images will have the HoK logo (maybe I will manually remove those logo and retrain them in future version, so try negative Logo & see if it helps).
100-128dim-64NA-512x
100 Steps - 128 Network Rank (Dimension) - 64 Network Alpha - 512 Max resolution
HoK-100-128dim-64NA-1024x
100 Steps - 128 Network Rank (Dimension) - 64 Network Alpha - 1024 Max resolution
Note: This model don't do too well on faces and male images probably because the training images didn't solely focus on faces and male images, so the faces is kinda weird sometimes, you've been warned.