Hey All, presenting my Daphne Blake LoRA from Scooby-Doo! Return to Zombie Island.<a target="_new" rel="ugc" href="https://en.wikipedia.org/wiki/Scooby-Doo!_Return_to_Zombie_Island">Scooby-Doo! Return to Zombie Island – Wikipedia</a><h3 id="base-models">Base Models</h3>Anima 1.0 — Base checkpoint (<code>anima-base-v1.0</code>). Not the Aesthetic or Turbo variants.Illustrious — Trained on <code>WAI-illustrious-SDXL_v11</code><h3 id="recommended-weight">Recommended Weight</h3>Anima<pre><code>0.9</code></pre>Illustrious<pre><code>0.8</code></pre><h3 id="trigger-word">Trigger Word</h3><pre><code>daphnertzi</code></pre><h3 id="style-(anima-and-illustrious)">Style (Anima and Illustrious)</h3><pre><code>2d animation, cel shading, flat colors</code></pre><h3 id="appearance-(anima-and-illustrious)">Appearance (Anima and Illustrious)</h3><pre><code>orange hair, long hair, purple headband, red lips, purple eyes</code></pre><h3 id="default-outfit">Default Outfit</h3>Anima<pre><code>green scarf around neck, purple dress, pink pantyhose</code></pre>Illustrious<pre><code>green scarf, purple dress, pink pantyhose</code></pre>For full-body generations, you can add:<pre><code>purple footwear</code></pre>The appearance and default outfit tags can be used separately depending on what you want to generate. The LoRA should also work with different outfits while retaining Daphne's identity.<h3 id="suggested-generation-settings">Suggested Generation Settings</h3>My testing was done using:ANIMA Settings with Turbo LoRA<ul><li>LoRA weight: 0.9 </li><li>Sampler: Euler a </li><li>Scheduler: Simple </li><li>Steps: 12 </li><li>CFG: 2.5 </li></ul>Illustrious SettingsUse appropriate quality and negative tags mentioned in the WAI_11 documentation, or you can use positive and negative embeddings. I generated my images using Smooth Positive and Negative embeddings along with Face Detailer for final face refinement.<ul><li>LoRA weight: 0.8 </li><li>Sampler: Euler a </li><li>Scheduler: Simple </li><li>Steps: 27 </li><li>CFG: 4.5 </li></ul>You can refer to my images for more examples. I have tried to tag the training images with accurate metadata and seeds.

Files with SHA256: 031a29bd35e695c1031cff07ae899ce22519d71ae6da9f78a57f48138649fa43