Type | |
Stats | 5,551 6,758 |
Reviews | |
Published | Feb 28, 2023 |
Base Model | |
Trigger Words | sonia \(pokemon\) |
Hash | AutoV2 A0EC9F2036 |
This is a micro-LoRA for Sonia from Pokémon Sword and Shield.
This was a REQUEST from one of the comments on my other pokemon LoRA.
I suggest using this LoRA at 0.75-1.0 weight for best results.
These images were generated and only tested using AOM3 A1 with the OrangeMix VAE and AnythingV3 VAE. (I'm lazy.)
Always add this to the prompt to activate it - Sonia \(pokemon\)
This LoRA works best with her traditional clothes - sunglasses, heart hair ornament, labcoat, green shirt, crop top, pants and other descriptive tags.
All pictures were upscaled using Hires. option on AUTO111
Upscaler: R-ESRGAN 4x+ Anime 6B
Hires. Steps: 5, Upscale by 1.5, Denoising Strength: 0.2
Known Issues
Sometimes (20% chance) It will generate two ponytails instead of one, you can decrease the chance by adding side ponytail to the prompt though.
If you don't add heart hair ornament - It will still generate but it will be warped.
Alot of her pictures on the internet and alot of the pictures in the dataset have her sunglasses super glued basically to her head, so even if you don't prompt for it - It will appear alot of the times.
Sometimes it's a little stiff but you can lower the weight of the LoRA to around 0.75-0.8 to make it a bit more flexible.
Embeds and LoRAs used in prompts - Bad Prompt - Extended Downblouse
Be sure to leave a comment, review and check out my other stuff here - you might like it.
Everything I'm making will always be free but if you want to show a little bit more support, you can buy me a coffee - https://ko-fi.com/ekune