This LoRa allows you to use Kevinsano's style.v3This version is trained at a resolution of 2048, you can see the difference in these images.v2 version<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/7f4ddc2e-1e71-4837-8dbb-1e9f44967f75/width=525/7f4ddc2e-1e71-4837-8dbb-1e9f44967f75.jpeg" />v3, same settings<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/1c113f60-eca9-41f5-9a9c-033dd40049aa/width=525/1c113f60-eca9-41f5-9a9c-033dd40049aa.jpeg" />I recommend using YiffMix, as it gives better results.v2I decided to look at what had gone wrong and it turns out that the images when resized to 1024x1024 lost a lot of quality. I reduced them to 2048x2048 manually and finally, the difference is noticeable and it looks better.I recommend using YiffMix or waiFuration , as it gives better results.v1Oh my god, this LoRa even using the same settings as the previous ones, always came out wrong, 6 versions and this is the sixth and last one.I'll put the dataset if someone wants to improve it, there are 150 images, all at 1024x1024But, they are all from MLP, for some reason of mine, I see them better than the others that are not from MLP.

Kevinsano Style

Files with SHA256: fe02aaca4cc464589d25f8e5e718ce7395901d8893b8990e1df7ab266a6c1948