<p>Hey All, presenting my <strong>Velma Dinkley LoRA</strong> from <em>Scooby-Doo! Return to Zombie Island</em>.</p><p><a target="_blank" rel="ugc" href="https://en.wikipedia.org/wiki/Scooby-Doo!_Return_to_Zombie_Island?utm_source=chatgpt.com">Scooby-Doo! Return to Zombie Island – Wikipedia</a></p><p></p><h3 id="update">UPDATE</h3><h3 id="krea-2-version">KREA 2 VERSION</h3><p>This is my <strong>second attempt at a Krea 2 LoRA</strong>. Due to hardware limitations, most of my testing was done using quantized models. I'm particularly interested in seeing how it performs on larger, higher-quality models, so please do share your generations and reviews!</p><h3 id="trigger-word">Trigger Word</h3><pre><code>velmartzi</code></pre><h3 id="recommended-weight">Recommended Weight</h3><pre><code>0.9</code></pre><h3 id="suggested-appearance-description">Suggested Appearance Description</h3><pre><code>an animated woman with a brown bob haircut and thick black-rimmed glasses. She wears an orange turtleneck sweater, a red pleated skirt, orange knee-high socks, and red Mary Jane shoes.</code></pre><p>You can take a look at my example images. One thing I noticed during testing is that <strong>scattering her default appearance across a natural-language prompt, rather than repeating everything in one rigid identity block, might give the model a little more freedom in how it interprets her appearance.</strong> Or it could just be my model!</p><p>Feel free to experiment with different prompt structures and outfits while retaining her recognisable appearance.</p><h3 id="suggested-generation-settings">Suggested Generation Settings</h3><p>My testing was done using:</p><ul><li><p><strong>LoRA weight:</strong> 0.9</p></li><li><p><strong>Sampler:</strong> <code>er_sde</code></p></li><li><p><strong>Scheduler:</strong> Simple</p></li><li><p><strong>Steps:</strong> 9</p></li><li><p><strong>CFG:</strong> 1.0</p></li></ul><p>I'd love to see what you create with either version, especially your results with Krea 2 on larger models. Please share your generations and leave a review. Happy generating!</p><p>_____________________________________________________________________________________________</p><h3 id="anima">ANIMA</h3><p><strong>Anima 1.0</strong> — Base checkpoint (<code>anima-base-v1.0</code>). <strong>Not</strong> the Aesthetic or Turbo variants.</p><h3 id="recommended-weight">Recommended Weight</h3><pre><code>0.9</code></pre><h3 id="trigger-word">Trigger Word</h3><pre><code>velmartzi</code></pre><h3 id="style">Style</h3><pre><code>2d animation, cel shading, flat colors</code></pre><h3 id="appearance">Appearance</h3><pre><code>short auburn hair, thick rectangular black-framed glasses, black eyes, freckles</code></pre><h3 id="clothes">Clothes</h3><pre><code>baggy orange turtleneck sweater, frilled red miniskirt, orange knee socks, red Mary Jane shoes</code></pre><p>The appearance and clothing tags can be used separately depending on what you want to generate. The LoRA should also work with different outfits while retaining Velma's identity.</p><h3 id="suggested-generation-settings">Suggested Generation Settings</h3><p>My testing was done using:</p><ul><li><p><strong>LoRA weight:</strong> 0.9</p></li><li><p><strong>Sampler:</strong> Euler a</p></li><li><p><strong>Scheduler:</strong> Simple</p></li><li><p><strong>Steps:</strong> 12</p></li><li><p><strong>CFG:</strong> 2.5</p></li></ul><p>You can refer to my images for more examples. I have tried to tag the training images with accurate metadata and seeds.</p>

Velma Dinkley (Scooby Doo - Return to Zombie Island)

<p>Krea2 Version 1.0 Velma Road to Zombie island</p>

Files with SHA256: 2b09bc74480a7cf2eb7b026aa684d22834aa4631d8fe3b0bc192e6bffb5e2d64