I love Graf Zeppelin, so I trained this model and use to collate and write the LoRA training document. You can also check <a target="_blank" rel="ugc" href="https://github.com/28598519a/lora-scripts-windows">https://github.com/28598519a/lora-scripts-windows</a>.I think the most difficult thing of train this LoRA is to have a good finger, iron cross, the symbols on the capelet at the same time.I've uploaded the preprocessed training data (located in the v1.0 option), thanks to the artists.Recommend to use Anythingv4.5 to generate anime-style images (AnythingV4.5 &gt; AOM2 &gt; AOM3 &gt; AnythingV3/V5、Counterfeit &gt; NAI)Usage:<ol><li>You can choose <code>graf zeppelin</code> or <code>graf zeppelin \(kancolle\)</code> as the main trigger word, but don’t use both (<code>graf zeppelin</code> is recommended)</li><li>The main style of Graf Zeppelin was split in 4 tags ( <code>uniform, capelet, hat, machinery</code> ), usually not used <code>machinery</code> (艤装 / ぎそう).</li><li>You can select some keywords to add in prompt e.g. ( <code>swimsuit, bikini, playboy bunny, maid, nun, japanese clothes, christmas, abyssal</code> ).</li><li>If you used v1.1 on a realistic model like chiloutmix-ni, the weight should give higher like &lt;lora:grafZeppelin_v11:1.2&gt;</li></ol>

Graf Zeppelin (kancolle / 艦これ)

Necktie seem to sometimes appear when they are not needed, but this version perform better than v1.0, fingers, background or something else.

Files with SHA256: 1aa832453fa3cb796e03d73e46b0465ab417e03323b61d0d3704b487da5c31b1