Verified: 7 months ago
SafeTensor
Type | |
Stats | 1,124 44,508 |
Reviews | (157) |
Published | Oct 9, 2023 |
Base Model | |
Training | Steps: 2,400 Epochs: 8 |
Usage Tips | Clip Skip: 2 |
Trigger Words | photorealism bear teddy_bear plushie |
Hash | AutoV2 4746EDC094 |
I originally ran this to teach myself a bit more about LoRA training settings. But it turned out good enough that I decided to upload it. It's trained on twelve images of a giant 50-inch tall teddy bear plush toy manufactured by a company called Sugar Tubs. But it can also be used to plushify characters by balancing the strength of their own LoRA against the weight of this one.
I've found that dialing down the strength of a character LoRA, but then using ((brackets)) or (emphasis:1.5) to crank up specific keywords for character details can help with getting a likeness nailed down.
It can also be used to generate non-bears, by just using plushie as a keyword, and leaving bear and teddy_bear out of your prompt.
Base Model is YiffyMix_V34