Weight : 1~1.2, Steps:25~30, CFG scale:4you can use natural language as promptsBeen super busy with real life lately, but a friend kept recommending Anima to me, so I finally gave it a shot.This is my first time training a LoRA for Anima, so I'm still figuring things out. But honestly, this model surprised me — the fact that it can run LoRA training locally without relying on the cloud is a huge deal for me. This round of training added natural language captions generated by Qwen. That said, my initial hyperparameters were just me winging it, and I only later came across <a target="_blank" rel="ugc" href="https://civarchive.com/models/2536147/greg-rutkowski-style-anima">the official Greg Rutkowski LoRA</a> training config as a reference. A lot of the settings can definitely be pushed more aggressively. So consider this just a starting point — more tuning to come.[<a target="_blank" rel="ugc" href="https://civarchive.com/models/2598040/">https://civarchive.com/models/2598040</a>]<h1 id="zoroj-:">zoroj :</h1><a target="_blank" rel="ugc" href="https://www.furaffinity.net/user/zoroj/">https://www.furaffinity.net/user/zoroj</a><a target="_blank" rel="ugc" href="https://twitter.com/ZoroJart"><s>https://twitter.com/ZoroJart</s></a><a target="_blank" rel="ugc" href="https://www.patreon.com/zoroj">https://www.patreon.com/zoroj</a><a target="_blank" rel="ugc" href="https://zoroj.tumblr.com/">https://zoroj.tumblr.com/</a><a target="_blank" rel="ugc" href="https://bsky.app/profile/zoroj.bsky.social">https://bsky.app/profile/zoroj.bsky.social</a><a target="_blank" rel="ugc" href="https://hyao.itch.io/lustful-desires">https://hyao.itch.io/lustful-desires</a><h2 id="prompts">Prompts</h2><ul><li>Prompt Prefix:</li></ul><pre><code>highres, absurdres,masterpiece, best quality, good quality,score_9, score_8, newest, recent, </code></pre><ul><li>Negative Prompt:</li></ul><pre><code>worst quality, low quality, score_1, score_2, score_3, artist name,blurry, jpeg artifacts, lowres, censor
</code></pre><h2></h2>

Files with SHA256: f2999e79450cbe422d2f72325bac8cf665732c4e31ed3e29e7ab7aa4bde72b2d