Type | |
Stats | 68 228 1.7k |
Reviews | (18) |
Published | Mar 13, 2025 |
Base Model | |
Training | Steps: 1,350 Epochs: 3 |
Usage Tips | Clip Skip: 1 |
Hash | AutoV2 039B020415 |
As a big alpaca fan, I couldn't resist creating my own alpaca LoRA π
In the research phase, I found the two cool alpaca LoRAs from @DerD4nnY, but I wanted a special kind of alpaca, the ones with extra long fur and extra fluffy. I also didn't intend to use super realistic real life photos like Danny did with his LoRAs, but also to use funny/odd AI-generated images of what I wanted the alpacas to look like.
So I decided to breed my own LoRA in the hopes of creating a new cool alpaca style π
For the training, I mainly followed this super helpful guide by @Dark_infinity. I selected 30 high-quality photos and self-created AI-generated images, which I cropped and retouched. Finally, I used JoyCaption to create captions, which I also edited manually. After weighing up the pros and cons, I have decided not to use any special trigger words. I don't think it makes sense in this case.
So the normal use of βalpaca(s)β in the prompt should be sufficient as a trigger.
As this is my first LoRA, I hope it works well. I'm open to any feedback.
You are also welcome to remix all the showcase images shown above.
May it deliver many amusing alpacas! π€π¦ππ