20 shown - 316 total matches
🚀 Run this workflow on Runninghub (Cloud GPU): https://www.runninghub.ai/post/2043609028336558081/?inviteCode=rh-v1497 (Claim 1000 FREE credits upo...
Create
Model repo on Hugging Face Hub.
<p>This V3.0 version was done the same way I've done my past models before I started playing around with the VAE. I hope you see more of a variety in characters, images are still clean, and still follows prompts well </p>
<p>This update brings further refinements across the board - better details and colours, improved consistency, and enhanced NSFW content.</p><p>This release comes with <strong>+100 new sample images</strong> added to the already extensive V3.0 gallery, covering an even wider range of styles and scenarios.</p>
Model repo on Hugging Face Hub.
<p>Changes</p><ul><li><p>Improved light handlings</p></li><li><p>Added some character / style knowledges</p></li><li><p>Handling both dynamic and non-dynamic posing</p></li></ul><p>NoobAI EPS v1.1 + Illustrious v2.0-stable DARE applied</p>
<p>Full BF16 checkpoint — 453 tensors, <strong>6.155B parameters</strong>, no NaN / no Inf. Clean build.</p><p>▸ <code>layers.*</code> (transformer blocks) — 390 tensors · <strong>5.43 B params</strong> · <strong>88.2%</strong> ▸ <code>noise_refiner</code> — 26 tensors · 361.8 M params · 5.9% ▸ <code>context_refiner</code> — 22 tensors · 353.9 M params · 5.8% ▸ <code>cap_embedder</code> — 3 tensors · 9.8 M params · 0.2% ▸ <code>final_layer</code> — 4 tensors · 1.2 M params · <0.1% ▸ <code>t_embedder</code> — 4 tensors · 0.5 M params · <0.1% ▸ <code>x_embedder</code> — 2 tensors · 0.25 M params · <0.1%</p><p><strong>Total: 453 tensors · 6.155 B parameters</strong></p><p>━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━</p><p>📊 <strong>Weight Distribution</strong></p><p>▸ <strong>Global range:</strong> min ≈ <strong>−14.00</strong>, max ≈ <strong>+13.94</strong> — in line with typical DiT checkpoints</p><p>▸ <strong>Most active modules</strong> (highest std): the deep layers <code>layers.26</code> through <code>layers.29</code> — this is where the style adjustment is concentrated. Their <code>ffn_norm2</code> and <code>attention_norm2</code> tensors show std up to <strong>3.2</strong> vs. a model average of ~<strong>0.32</strong></p><p>▸ <strong>Most conservative modules:</strong> <code>t_embedder</code> (std ~0.005–0.02) — timestep embedding is nearly untouched, as expected from a weight-adjustment rather than a retrain</p><p>▸ <strong>Feed-forward </strong><code>w2</code><strong> weights</strong> carry the largest absolute values (up to ±14), consistent with how Z-Image's...
🚀 Run this workflow on Runninghub (Cloud GPU): https://www.runninghub.ai/post/2043861564398247937/?inviteCode=rh-v1497 (Sign up via the link to get...
Create
<p><strong>SatyrRemix - Iris</strong></p><p><em>Stylized Series</em></p><p>The second entry in SatyrRemix's Stylized Series — a watercolor-based SDXL merge that trades Lycoris's nostalgic softness for bolder color and brighter saturation.</p><p>Iris shares the same foundation as Lycoris: that hand-painted watercolor texture, the gentle color bleeds, the illustrated-world feel. But where Lycoris leans muted and storybook-warm, Iris turns the dial up. Autumn foliage burns deeper orange, skies stretch into clearer blues, character hair catches jewel-tone highlights, and backgrounds pop with saturated depth. Think less "aged storybook," more "freshly inked art print."</p><p><strong>What it does well:</strong></p><p>Iris is built for subjects that benefit from vivid color work — Art Nouveau portraits with rich floral palettes, sweeping fantasy landscapes with golden sunsets over turquoise lakes, character designs where hair and costume colors need to sing. The watercolor texture keeps everything feeling painted rather than digital, but the boosted saturation gives your generations more visual punch and energy. Cozy tavern scenes still feel cozy; they just glow a little brighter now.</p><p><strong>What to keep in mind:</strong></p><p>Same rules as the rest of the Stylized Series — the aesthetic is baked in, artist tags will blend with the base look rather than override it. If you want the softer, more muted storybook feel, reach for <strong>Lycoris</strong>. If you want full style control, go to <strong>Ankara</strong>. Iris is for when you want your painterly world turned up...
Source context: 0 downloads · Type Checkpoint · Base model Illustrious
Model repo on Hugging Face Hub.
<p>A major update enhancing lighting effects and stylized content. Side effects are temporarily unknown. If you encounter any issues, feedback is welcome!</p>
<h1 id="unstable-revolution-zit"><span style="color:rgb(250, 222, 243)"><strong><em>unStable Revolution ZIT</em></strong></span></h1><p><strong>Z-Image Turbo, improved with realism and NSFW</strong></p><h3 id="recommended-settings:"><u>Recommended Settings:</u></h3><h3 id="v3">v3</h3><p>Improved everything</p><p>Sampler: Euler</p><p>Scheduler: Simple</p><...
<p>WildCanvas-v0.1 is a merged image model crafted for furry art with an emphasis on feral and wild aesthetics. Human hairstyles are intentionally excluded to keep characters feeling true to their animal nature.</p>
<p>EH Cartoon 4.0 has renewed western style aesthetics. It's completely different from 3.0 (if you're looking for more anime aesthetics 3.0 or 2.0 is a better choice). </p>
Source context: 4 downloads · Type Checkpoint · Base model Illustrious
<p>Merged AnimaIka, animaCatTower, AnimaYume, and several other LoRAs.</p><p>This has reduced the variety of low-quality seeds.</p>
<p><span style="background-color:rgb(26, 27, 30);color:rgb(230, 73, 128);font-family:-apple-system, BlinkMacSystemFont, "Segoe UI", Roboto, Helvetica, Arial, sans-serif, "Apple Color Emoji", "Segoe UI Emoji";font-size:16px">Wan 2.7 will become available on the Civitai Generator soon!</span></p><p></p><p>Original details from:...
<p>And here we go again, after more than 300 iterations I've come to this new version.</p><p>I have some <strong>BAD news</strong>:</p><ul><li><p>this version has been fine-tuned on ComfyUI, which means that the results on Forge will be less optimized - basically worse than the 3.0 version.</p></li><li><p>this version is comparatively less reliable than the 3.0, if you want consistent results, you might want to try that previous version.</p></li></ul><p>Let's come to the <strong>GOOD news</strong>!</p><ul><li><p>Version 3.5 really unlock the potential to use dramatic perspectives and move the camera a lot more than the previous versions. This sometime brings to some weirdness, but it's really worth it to play with the keyword "Perspective" now.</p></li><li><p>Cleaner lines than ever before and pretty good at doing hands!</p></li><li><p>Large variety of styles staying inside the "2D Western" aesthetic</p></li></ul>
<p>Focused on improving contrast, darker darks and lighter lights, as well as higher resolution stability allowing for 1.5 megapixel base gens without anatomy issues in most cases and sometimes up to 2 megapixels (this also means hires should be more stable as well).<br /><br />Tried to improve overall details and creativity of the model while keeping its prompt adherence and knowledge, unfortunately I did not have the time to train in new characters or artists and merging models that had that knowledge was not as reliable as I had hoped.</p>
Source context: 1265 downloads · Type Checkpoint · Base model Illustrious
<h1 id="cinematic-photorealism-turbo-checkpoint"><strong>Cinematic Photorealism Turbo Checkpoint</strong></h1><p><strong>Ultra-realistic generation optimized for speed, precision, and material fidelity.</strong><br />Delivers professional-grade photography results in as few as <code>9 steps</code>. Fully compatible with FP8 quantization.</p><hr /><h2 id="...
<p>- More realism.</p><p>- Much better NSFW content.</p><p>- Prompts are better respected</p>
Source context: 1299 downloads · Type Checkpoint · Base model ZImageTurbo
