<a target="_blank" rel="ugc" href="https://ko-fi.com/hat0ri"><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/7a1e5f30-004e-4d68-90d3-adca38b27ea5/original=true/7a1e5f30-004e-4d68-90d3-adca38b27ea5.jpeg" /></a><h3 id="general:">General:</h3>This model is giving me a headache to be honest - it's a mess. There comes a time, when the amount of trained versions is quite large and working through them becomes a huge chore... It's still work-in-progress, but I just need to upload something, so as to not quit.I've tried many things, the pixels don't come out right either way. That's probably mostly due to the VAE, since I've tried everything I could for the training to not mess the pixels up.Also, one thing to note is that this was trained on the preview3 version, but I'll be making a full switch to the Anima-base 1.0 version from now on, so I generated the previews with it and the lora works well still.For the pixels it's definitely advisable to scale the images down with nearest neighbor and quantize them to fix the pixel grid.<h3 id="tags-and-information">Tags and information</h3>Use tags:@l0ph1ro, pixel art+ optionally two character appearances<ul><li>rytsar, 1boy, knight, helmet, plume, short hair, brown hair, chainmail, pauldrons, armor</li><li>p0ryot4, 1girl, freckles, blue eyes, blonde hair, single hair bun, grey scarf</li></ul>and some other appearances of the creatures from the game have also been learned (not a focus, so not exact appearances - see preview images)*. An issue I noticed is that landscapes tend to generate a border (that's due to padding during rescaling to fit buckets - I'll probably just give up on this approach, so this will be fixed)Support:If you like what you find here, please consider making a donation <a target="_blank" rel="ugc" href="https://ko-fi.com/hat0ri">https://ko-fi.com/hat0ri</a>, thanks!

Files with SHA256: 2635e828e649c85be9453cd93c7e531e3ba5cc068f101f3046624f018d9e448d