Type | |
Stats | 102 |
Reviews | (16) |
Published | Feb 3, 2025 |
Base Model | |
Hash | AutoV2 4A8F101A49 |
Ana De Armas, a Cuban and Spanish actress.
I continued my experiment and ended up with this 1.5MB LoRA.
Even though it's just 1.5MB, the results are actually pretty decent.
Here's what I noticed when testing it for image generation:
1. The image is very clean.
2. Since it's so small, pretty much all the details like clothing, backgrounds, and art styles from the training dataset got tossed out. Apart from the target character, this LoRA has very little impact on the generated images.
3. The stability isn't the best—there's a chance it might spit out results that don't really look much like the target person. And that's pretty much tied to what's in the prompt.
4. It works pretty well alongside other LoRAs, but it can easily get influenced by them.
If I keep tweaking the parameters, I could probably shrink the LoRA even more, but honestly, I don't think it's worth the effort at this point.
Wish you hace fun!
TI(Embedding) version: https://civitai.com/models/559829/ana-de-armas-soloti