<h3 id="atomix-xl-v4.0-tiq89teu3"><span style="color:rgb(250, 176, 5)">Atomix XL V4.0</span></h3><p>improved the understanding of prompt.</p><p></p><h3 id="atomix-xl-v3.0-kibuv4tjv"><span style="color:rgb(250, 176, 5)">Atomix XL V3.0</span></h3><p>Still a Lightning version,it is base on V2,improved the understanding of prompt and image quality.</p><p></p><h3 id="atomix-xl-v2.0-eegb2fce9"><span style="color:rgb(250, 176, 5)">Atomix XL V2.0</span></h3><p>V2.0 is a Lightning version, its generation speed and image quality are much better than V1.0.</p><p></p><h3 id="atomix-xl-v1.0-65bka7a1s"><span style="color:rgb(250, 176, 5)">Atomix XL V1.0</span></h3><p>I only used the SD1.5 model before,not only because the generation speed of the SDXL model is too slow, but also because the generated image are too realistic(too many film grain).In the past two days, I occasionally tried SDXL Turbo model and was surprised to find that the speed of SDXL has greatly improved, the time spent is completely acceptable. So I started trying on solving the rest problem, and I succeeded. So, this is a SDXL model that balances reality and CG well,which generate clean images(fewer film grain),and supports Turbo acceleration.</p><p></p><h3 id="recommended-settings-hhk9osqus"><span style="color:rgb(250, 176, 5)">Recommended settings</span></h3><pre><code>cfg: 1~2 
steps: 6~10 
sampler: dpmpp_sde_gpu
scheduler: sgm_uniform</code></pre>

Atomix XL

atomixXL_v20.safetensors

Files with SHA256: cd6beca4e6803e4dc20c6f51c24857ab2340d3e0085d6243a0afecba879360c1