Please read through the entire description (might need to be expanded) and the version change notes as they cover a lot of information about basic use cases and limitations. The description is always focused on the latest version. Thank you!This is an SD 1.5 LoRA for the character May / Haruka from Pokemon. It works for models based on SD 1.x only, it will not work for models based on SD 2.x or SDXL!Example images were picked from pure 512x512 txt2img results and then re-created at 1024x1024 using txt2img with moderate Hi-res Fix settings (upscaler Lanczos, denoising strength 0.4). This improves faces and other details that are almost impossible to get correctly and consistently at 512x512 (limitation of the technology) while still giving a realistic impression of what it looks like. Pure 512x512 results will have more distorted faces and less detail.Please see the version change notes for the training and example image generation models as well as the used weights as they might change between versions. Remember that you might need to adjust weights to best suit your use case!Remember to add the tag may \(pokemon\) (with backslashes intact) to your positive prompt.The training set contained multiple variations of May's signature look (not including the ones from Emerald and ORAS, sorry) so you might need to put combinations of the following tags in your positive or negative prompts to get the desired results. Some were included with both more generic and specific tags (separated by slashes in this list) and those might need to be combined or used on their own:<ul><li>bandana</li><li>bike shorts</li><li>fanny pack</li><li>gloves</li><li>shirt</li><li>shoes</li><li>skirt</li><li>socks</li></ul>As the training set contained a few images with floating hair, you might need to add floating hair to the negative prompts if not desired.Known Limitations / Problems:<ul><li>The shoes were not present in a lot of images and not a focus for the training so they sadly will not be consistent at all.</li><li>Same goes for the fanny pack and the pattern on the bandana (though they tend to be a bit better).</li><li>The colors of clothing and accessories can be all over the place, sometimes even swapping around. Especially for closer shots and from behind. Don't really know why yet, I might have to tinker with the training images and/or tags a bit more. Maybe I might have to use explicit color tags in the future. It might also help to just put the colors in the positive prompts for now.</li><li>It can be a bit difficult to get images without the bandana or any other headwear, depending on the remaining clothing and framing. In addition to bandana, I would also recommend additional negative prompts such as hat. Maybe even put no headwear into the positive prompts.</li><li>It seems very difficult to remove the shirt on its own while keeping the skirt and bike shorts by just using negative prompts. I've tried a lot of things during training but it still does not seem to work consistently. Try adding something like bra or topless to the positive prompts and shirt to the negative prompts and maybe you will get lucky.</li><li>Might generate a collar even when the shirt is not present. Adding collar to the negative prompt might help but I've not found a consistent fix yet.</li><li>The hairstyle from the side and from behind is very inconsistent and might not look correct due to very limited training data at the moment.</li><li>The entire appearance (especially hairstyle and clothing) will probably look off from behind due to even more limited training data.</li><li>The bandana shape, separate skirt and glove coloring in particular will probably take a lot of tries to get looking OK from any perspective. They seem to be pretty difficult to replicate (especially together) or are still missing good training data.</li></ul>

[SD 1.5] Pokemon - May / Haruka

There are detailed changes below the next paragraph, you might need to expand this version changes box!I learned a few new things about creating LoRAs and wanted to see how far I can push my existing LoRAs without changes to their datasets. So there's still no support for alternate outfits, colors etc. for now, sorry! There are other LoRAs if you want those things. Not sure if I'll improve the LoRAs more than that in the future or make new ones but who knows.LoRA Changes:<ul><li>Now uses the standard "may \(pokemon\)" instead of a custom trigger word to make things easier for everyone (please note though that the LoRA itself is still named "character_pokemon_may_vX" as before!)</li><li>Switch to 32 dims (now trained directly instead of resizing from 128)</li><li>Switch to target weight 1 (increased from 0.7)</li><li>Changed a lot of training settings (too many to list, check LoRA metadata if you are interested) and updated to much newer scripts version (~3 months worth of changes)</li></ul>Example image changes:<ul><li>Use CLIP skip 2 instead of 1 as I accidentally did before</li><li>Change settings for hir-res fix: Switch upscaler from None to Lanczos and increase denoise from 0.1 to 0.4 to get better results while still not changing TOO much</li><li>Use a negative embedding for the majority of negative prompting</li></ul>Recommended weight: 1Training model: <a target="_blank" rel="ugc" href="https://civitai.com/models/66/anything-v3">Anything V3</a>Example image generation model: <a target="_blank" rel="ugc" href="https://civitai.com/models/4451/abyssorangemix2-hardcore">AbyssOrangeMix2 - Hardcore</a>

Files with SHA256: d6302eabdef333579b52c879a3c4174ca11e9bbc04c15350f61d35fb80886ce5