<h1 id="katsura-masakazu-style-7myua4zc0">Katsura Masakazu Style</h1>Reviews are very appreciated and show me you enjoy my work, keeping me motivated in releasing more Models in the future! Would love to see the community create images with my work.Because of the low Dataset on danbooru the base Illustrious/NoobAI Models can't really do this artists style very well.Best Results are at 1 weight &amp; using the trigger word:<pre><code>Katsura Masakazu</code></pre><h2 id="v1.0-(noobai-eps)-9rv42poee">v1.0 (NoobAI EPS)</h2>Trained on the base <a target="_blank" rel="ugc" href="https://civarchive.com/models/833294?modelVersionId=998979">NoobAI 0.75 EPS Version</a> model for compatibility with all other NoobAI models. I chose 0.75 because of lower saturation of colors from the base model.All preview images are created using the <a target="_blank" rel="ugc" href="https://civarchive.com/models/913998?modelVersionId=1122850">NoobAICyberFix 3.0</a> Checkpoint.Best Results are at 0.8 weight &amp; using the trigger word:<pre><code>Katsura Masakazu</code></pre>The settings used to create the images were:<pre><code>Steps: 30
CFG: 6
Sampler: Euler
Scheduler: SGM_Uniform
Denoise: 1</code></pre><h2 id="v1.0-(il)-rdbj101ji">v1.0 (IL)</h2>Trained on the base <a target="_blank" rel="ugc" href="https://civarchive.com/models/1232765/illustrious-xl-10">Illustrious XL 1.0</a> model for compatibility with all other Illustrious models.You can see a comparison between just the tag itself and the influence with the LoRA <a target="_blank" rel="ugc" href="https://i.imgur.com/CtKWxxW.jpeg">here</a>. The difference is huge.All preview images are created using the <a target="_blank" rel="ugc" href="https://civarchive.com/models/833294?modelVersionId=1116447">NoobAI-XL Epsilon-pred 1.1-Version</a>. The comparison images in the table were made by using the same seed and settings with different prompts using the normal <a target="_blank" rel="ugc" href="https://civarchive.com/models/795765?modelVersionId=889818">Illustrious-XL Model</a>.Best Results are at 1 weight &amp; using the trigger word:<pre><code>Katsura Masakazu</code></pre>The settings used to create the images were:<pre><code>Steps: 30
CFG: 5
Sampler: Euler
Scheduler: Normal
Denoise: 1</code></pre>All images were upscaled using the <a target="_blank" rel="ugc" href="https://openmodeldb.info/models/4x-Remacri">Remacri-4x Model</a>. All images were generated using my ComfyUI workflow that you can find <a target="_blank" rel="ugc" href="https://civarchive.com/models/1297813/comfyui-detaileradetailer-all-in-one-illustrious-pony-sdxl">here</a>, attached to preview images or in the suggested resources. This includes a Detailer that automatically inpaints body parts like eyes, lips, hands etc. So if your results do not match mine it's probably because of this. I recommend using the workflow which is also attached to every image.

Katsura Masakazu - Retro Manga / Anime Style [NoobAI, Illustrious]

<ul><li>Trained on <a rel="ugc" href="https://civitai.com/models/833294?modelVersionId=998979">NoobAI epsilon prediction model v0.75</a> (has lower color saturation) and therefore closer to artists depiction</li><li>Way better anatomy and prompt adherence than the first IL version</li><li>Trained on 1024x1024 for better quality and anatomy on eyes and hands especially</li></ul>

KatsuraMasakazu.safetensors

v1.0 (NoobAI EPS)

v1.0 (IL)

<h1 id="katsura-masakazu-style-7myua4zc0">Katsura Masakazu Style</h1>Reviews are very appreciated and show me you enjoy my work, keeping me motivated in releasing more Models in the future! Would love to see the community create images with my work.Because of the low Dataset on danbooru the base Illustrious/NoobAI Models can't really do this artists style very well.Best Results are at 1 weight &amp; using the trigger word:<pre><code>Katsura Masakazu</code></pre><h2 id="v1.0-(noobai-eps)-9rv42poee">v1.0 (NoobAI EPS)</h2>Trained on the base <a target="_blank" rel="ugc" href="https://civitai.com/models/833294?modelVersionId=998979">NoobAI 0.75 EPS Version</a> model for compatibility with all other NoobAI models. I chose 0.75 because of lower saturation of colors from the base model.All preview images are created using the <a target="_blank" rel="ugc" href="https://civitai.com/models/913998?modelVersionId=1122850">NoobAICyberFix 3.0</a> Checkpoint.Best Results are at 0.8 weight &amp; using the trigger word:<pre><code>Katsura Masakazu</code></pre>The settings used to create the images were:<pre><code>Steps: 30
CFG: 6
Sampler: Euler
Scheduler: SGM_Uniform
Denoise: 1</code></pre><h2 id="v1.0-(il)-rdbj101ji">v1.0 (IL)</h2>Trained on the base <a target="_blank" rel="ugc" href="https://civitai.com/models/1232765/illustrious-xl-10">Illustrious XL 1.0</a> model for compatibility with all other Illustrious models.You can see a comparison between just the tag itself and the influence with the LoRA <a target="_blank" rel="ugc" href="https://i.imgur.com/CtKWxxW.jpeg">here</a>. The difference is huge.All preview images are created using the <a target="_blank" rel="ugc" href="https://civitai.com/models/833294?modelVersionId=1116447">NoobAI-XL Epsilon-pred 1.1-Version</a>. The comparison images in the table were made by using the same seed and settings with different prompts using the normal <a target="_blank" rel="ugc" href="https://civitai.com/models/795765?modelVersionId=889818">Illustrious-XL Model</a>.Best Results are at 1 weight &amp; using the trigger word:<pre><code>Katsura Masakazu</code></pre>The settings used to create the images were:<pre><code>Steps: 30
CFG: 5
Sampler: Euler
Scheduler: Normal
Denoise: 1</code></pre>All images were upscaled using the <a target="_blank" rel="ugc" href="https://openmodeldb.info/models/4x-Remacri">Remacri-4x Model</a>. All images were generated using my ComfyUI workflow that you can find <a target="_blank" rel="ugc" href="https://civitai.com/models/1297813/comfyui-detaileradetailer-all-in-one-illustrious-pony-sdxl">here</a>, attached to preview images or in the suggested resources. This includes a Detailer that automatically inpaints body parts like eyes, lips, hands etc. So if your results do not match mine it's probably because of this. I recommend using the workflow which is also attached to every image.

Files with SHA256: f06cff486c78eea8d6d6b89dd082c382a4558ca0493bcf8f1915c52882f30687