Type | |
Stats | 5,274 43,756 |
Reviews | |
Published | May 16, 2023 |
Base Model | |
Trigger Words | sharp teeth |
Hash | AutoV2 D9E88C6B0C |
I'm gonna be real with you:
I have no idea if this actually improves anything or not. Of course, it was trained in the hopes of making sharp teeth more neat and clean, but as far as that goes, I'm not sure if I succeeded. So any feedback would be appreciated so I can know for sure if this actually helps. From my tests, even without the lora, it seems like models on their own can manage somewhat fine without this.
You can check out the previews above. Portrait style generations are...hard to tell the difference with. If anything, this lora seems to maybe add more teeth than those that generated without. Face details naturally get muddier the farther the subject is, but with this, it seems like it helps slightly to make sharp teeth a bit more coherent from a distance. But again, the differences are very minmal, so perhaps it's best tested through situation-cased basis. I had charts to distinguish the images but deleted them because originally I was going to scrap this.
I can't recommend any weight settings, because I'm not sure at what weight it's most effective at.
Discussion
Regarding: "I have no idea if this actually improves anything or not."
You can do an X/Y-plot that shows without LoRA and with LoRA at different weights, that way you can easily show what it does and at what weights.
Those plots are of great help for us users too :)
