Type | |
Stats | 312 1,745 555 |
Reviews | (90) |
Published | Apr 9, 2025 |
Base Model | |
Training | Steps: 1,103 Epochs: 15 |
Usage Tips | Clip Skip: 2 Strength: 0.9 |
Trigger Words | dsny2 |
Hash | AutoV2 BB2055DF8C |
Done a few tests myself, and I'm pretty happy with this version. This LoRA was trained on generated images that resembles a more Classic Disney style.
Recommended
Strength 0.9, lower for more flexibility.
CFG 3.5 or lower, anything above that will start giving artifacts and blow out the image.
Only trigger word needed is: dsny2
Thank you again for using my LoRA, you can support me through either donating Buzz, or you can use one of the links provided beneath my profile.
Changelog: 8/4/2025
Retrained the entire dataset with a more focused set of images that represents the mix without changing /too much/ the overall original style (I had some images with different qualities and overall 'too different' style from the Disney style, that it ended up mixing a bit too much on the first iteration). This time, though, the style is a lot of more consistent than before.
Different settings used on the training aswell, which focused more on keeping a more consistent style overall.
Fixed the dataset overall.
Hope this new version is more consistent, I've been testing it out for a few hours now and I gotta say, it does seem consistent. The images I'm posting right now shows that, they're literally the same settings (mostly) and I kept the original positive prompt and just changed the rest of the prompt for different results, let me know what you guys think.