Verified: a year ago
SafeTensor
Type | |
Stats | 59 0 |
Reviews | (3) |
Published | May 28, 2024 |
Base Model | |
Training | Steps: 38,100 Epochs: 1 |
Trigger Words | ELS style |
Hash | AutoV2 7FD12C4D1A |
My second attempt at creating my own LORA.
Style of Everlasting Summer, a Russian 2013 visual novel video game. There are a lot of LORAs dedicated to certain characters from the game, but I wanted something representing its style.
381 images were used to train the model. It included characters, landscapes, scenes.
The training of the LoRA was on stable-diffusion-v1-5 model.
Max resolution was 1024x1024.
I recommend to set the strength of this LORA to 0.5-0.7. It has quite good results even without trigger words 'ELS style'. Their inclusion can drastically change the image, though.