PLEASE DO NOT REPOST MY MODEL ON OTHER SITES.A monomasa LoRA trained on <a target="_blank" rel="ugc" href="https://civarchive.com/models/92450">FluffyRock, </a>for the sidekick form only.Manually tagged dataset, clothing tags used:<pre><code>white hooded coat,
hood up,
hood down,
yellow belt,
black pants,
black tank top,
necklace,</code></pre><h2 id="heading-117">V1.2</h2>The initial intention of this version was to enhance the precision of necklace modeling, but there was no significant improvement observed after training. However, there might be other improvements that I haven't noticed. Perhaps you could try and see for yourself.Tags not changed since v1.1.<h2 id="heading-2186">V1.1</h2>Slightly improve the eye details, but might be harder to control.Some tags are updated since they are inaccurate in the last version:<pre><code>detailed eyes (add this but not at the very beginning)
black pupils, green eyes, red sclera, 
yellow eyeshadow (less than 0.8 is better)
thick eyebrow (optional)</code></pre>I suggest use the UniPC(text2img) + Restart(hi-res fix) + R-ESPRGAN 4x+Anime6B upscaler with 0.47+ Denoising strength to get best result.You can try use the Region prompt extension on webui to make it better.<h2 id="heading-2187">V1.0</h2>You may also need to add<pre><code>green pupils, red sclera/colored sclera for eyes,
wolf/husky(even he is not, lol) for species,</code></pre>to get best results.==================================================Showcase images are generated using <a target="_blank" rel="ugc" href="https://civarchive.com/models/34469">IndigoFurryMix</a>Add proper artist tags to achieve your favorite styles. It is very important.==================================================! May generate NSFW content if you not tagged the clothing properly.==================================================

[LAH] Monomasa Sidekick

A new version, trained with the new Prodigy optimizer. Implemented changes to reduce repetition and increased the number of epochs to achieve the total steps needed.This update results in generating different face shapes compared to the v1.1 version.Note: This version's dataset includes some AI-generated images.

Files with SHA256: b291386fa182444799a18c0035ea8441ce054dc19820666b37524a6750a5ac35