<h2 id="training-information">Training Information</h2>Trained in <a target="_blank" rel="ugc" href="https://github.com/ThetaCursed/Anima-TrainFlow">Anima-TrainFlow</a><ul><li>This LoRA is trained using Anima Base with the total image dataset of 56+ which combines Antal from Arknights Endfield and 12F from Arknights.</li></ul><h3 id="promptparameter-settings:">Prompt/Parameter Settings:</h3>Lora Strength: .7 - .9<h2 id="base">Base</h2>Activation Tag:<pre><code>1boy, antal (arknights), furry</code></pre>Reinforcement Tags:<pre><code>dragon boy, tail, yellow eyes</code></pre><h2 id="outfit">Outfit</h2>12F's Attire:<pre><code>black coat, hood, hooded coat, black pants, long sleeves</code></pre>Antal's Attire:<pre><code>black pants, black shirts, white cropped jacket, gloves, high collar,knee pads,</code></pre><h3 id="after-thoughts">After thoughts</h3><ul><li>The final 24k step for this lora is just downright horny without even me prompting something NSFW. That why I settled for 21K Step instead so people who are generating this lora wouldn't get flashed with unintentional NSFW generation. Not to mention, there will be some generation where 21k Step Lora would somewhat generate his head accurately compare to the final 24k Step Lora when it comes to his head when generating with artist tags.</li><li>Nonetheless, It can accurately generate 12F's attire except the Antal's Attire, It somewhat captures his attire but is losing the extra details on his collar and is adding extra details on his cropped jacket which doesn't exist in the official design.</li></ul><s>Anyways, this lora should atleast be able to generate Antal consistently. I wasn't really able to generate him clearly when using some other Arknight Endfield Character Pack Lora.</s>

Files with SHA256: a4fdfb89c01c1ebf3f131149f307adf2c3ff4e6b26213712947eaa446b74ea6e