Hassaku aims to be a anime model with a bright and distinct anime style.My <a target="_blank" rel="ugc" href="https://discord.gg/zSR5FcYWWE">Discord</a> for everything related to anime models and art. If you'd like to support my work, you can do so at: <a target="_new" rel="ugc" href="https://subscribestar.adult/citrus-models">SubscribeStar</a>, every bit helps and is truly appreciated! ____________________________________________________________ Supporters:Thanks to my supporters tamashiicolle, pttcot, SETI and Kodokuna ____________________________________________________________ Using the model:<ul><li>The model was trained primarily on images with minimal disruptive elements, such as floating text, logos, speech bubbles, or signatures. If any of these elements appear in an image, please include "signature" in your negative prompt.</li><li>Danbooru <a target="_blank" rel="ugc" href="https://danbooru.donmai.us/wiki_pages/tag_group%3Ametatags">Metadata</a> and franchise tags are not used in training, so please avoid using them in prompts. Metadata such as "highres" and franchise tags like "re:zero kara hajimeru isekai seikatsu" are not used in training.</li><li>A CFG scale between 3 and 7 is recommended. For best results, use CFG 6 with the Euler a sampler.</li><li>Refer to the example images, to understand, how the model should be used.</li><li>The model is designed to be simple and straightforward to prompt. Use the following tag order:<ol><li>Number of characters</li><li>Character Tags</li><li>All remaining tags Example: 1girl, rem \(re:zero\), standing, masterpiece, upper body</li></ol></li><li>Here are some resolution options for SDXL:<ul><li>1536 x 640</li><li>1344 x 768</li><li>1216 x 832</li><li>1152 x 896</li><li>1024 x 1024</li><li>896 x 1152</li><li>832 x 1216 (most recommended)</li><li>768 x 1344</li><li>640 x 1536</li></ul></li></ul>Lora use:<ul><li>NoobAIs Loras working for the most part better in compare to <a target="_blank" rel="ugc" href="https://civarchive.com/models/795765/illustrious-xl">Illustrious loras</a></li><li>Versions 2 and 3 is less compatible to most loras trained on Illustrious or Noob, because it got further trained away of both. Style A is still the most compatible for it.</li></ul> ______________________________________________________Version and License info:<ul><li>Below V1 merges <a target="_blank" rel="ugc" href="https://civarchive.com/models/260267">ANIMAGINE XL 3.0</a></li><li>Version V1 use <a target="_blank" rel="ugc" href="https://civarchive.com/models/795765/illustrious-xl">Illustrious-XL</a> &amp;<a target="_blank" rel="ugc" href="https://civarchive.com/models/827184?modelVersionId=1068947"> WAI-NSFW-illustrious-SDXL</a> with additional training</li><li>V2 is trained on its own and don't include any extra merge. Base was V1, it is trained to include newer or missing characters. It was also used for further training tests</li><li>V3 is a merge of <a target="_blank" rel="ugc" href="https://civarchive.com/models/795765/illustrious-xl">Illustrious-XL</a> and V2.2, to fix some issues and was also trained to include newer or missing charactersAll Models using the <a target="_blank" rel="ugc" href="https://freedevproject.org/faipl-1.0-sd/">Fair AI Public License 1.0-SD</a> license.</li></ul>

Hassaku XL (Illustrious)

Update, mainly focused on fixing issues from v3.3 Better: - Fewer artifacts and less face/eye melting, especially at a distance - Cleaner, less blurry outputs - Additional training on the same dataset af v3.3, some characters working now better What i noticed in 3.3 + now 3.4: - Hair color prompts are less reliable (eye colors gets used for the hair, as example) - Skin can appear extremely bright/shiny again - Multiple views occurs more often I believe these issues started after training with Prodigy, it is maybe to agressive. I’ve since switched to ADOPT, and results improved a bit. I may release another update with further fixes and an updated dataset. Problem is that I’m still unsure when Z-Image or Klein will get a proper anime model. I also have several experimental datasets I want to test train, including mixes of natural language and tags (including NSFW, tags combine with natural language, that describe spatial awareness of characters and positions of texts, the exact weakness of tag based models).

Files with SHA256: 1618edb443d9c641fb01c4961f5875d5f01b1a851fd9f2ea64623ceac82257e0