Here i will upload 3d and 2.5d models to keep Mergestein Animu only 2D.Available on Pixai for on-site generation (NSFW Works):<ul><li><a target="_blank" rel="ugc" href="https://pixai.art/es/model/1970418852754204169/1970418860308146223">2.5D A+ Reforge</a></li><li><a target="_blank" rel="ugc" href="https://pixai.art/es/model/1962218204941519374/1962218206145284760">3D A+</a></li><li><a target="_blank" rel="ugc" href="https://pixai.art/es/model/1961807929686636468/1961903406620240071">Photorealistic 3D</a></li><li><a target="_blank" rel="ugc" href="https://pixai.art/es/model/1995988579365420211">Simple Cuteness Enhanced</a></li></ul>Simple Cuteness Enhanced: Same as my other version but with LoRas merged in, if you are lazy about LoRas, this has more details, if you like to use detail LoRas, the other one is a lot better. The model has a CLIP merge of just WAI 16, Janku 6.9, Chenkin 0.2 and NoobAI 1.1, so it should have good prompt adherence.2.5D A+ Reforge: Merge of WxJ+NM-PM Reforge with Mergestein Animu Plus Alter and WxJ+NM-PM v5.5.3D A+: 2.5D-3D but keeps anime/toon faces.Photorealistic 3D: Merge result of my Mergestein Animu-A Plus Alter and an updated version of my Mergestein Uncanny-A Plus Realistic (not released). As a new model cuz is not realistic enough to be on my Uncanny Realism model (still if you prompt realistic, photorealistic it can do that so, it's just the base style that is 3D). Due to being less realistic, unrealistic prompts work a lot better like motion lines, speed lines, sound effects, etc. The model has merged in <a target="_blank" rel="ugc" href="https://civarchive.com/models/152040?modelVersionId=1191929">XL_VAE_C - G9</a> that add a lot of contrast, so it can look too dark, if you want to remove it, use this VAE: <a target="_blank" rel="ugc" href="https://civarchive.com/models/152040?modelVersionId=1803891">XL_VAE_C - F2-3.1 </a>or this one: <a target="_blank" rel="ugc" href="https://civarchive.com/models/296576/sdxl-vae">SDXL VAE. </a>This will make the images looks softer and less dark.The cover image of photorealistic 3D is censored to make the model shows to more people while browsing, but it can do NSFW.You can use Illustrious, SDXL and some pony LoRas, the quality can vary due to different base training model, so you need to test by yourself.The model uses booru tag system (1girl/1boy), but you can use descriptive prompt too.The images i upload use Euler Beta and without using any CFG Rescale, Hiresfix, ADetailer or Upscale. So, the quality can be improved. Close distance is perfect; medium is ok and far is ugly.Samplers and Schedulers:<pre><code>Euler a Normal/Simple/SGM Uniform 2-4 CFG for general use</code></pre><pre><code>Euler Beta/SGM Uniform 2-4 CFG for sharper images, recommended </code></pre><pre><code>DPM++ 2/3/4M SDE Karras 1.5-3 CFG not recommended, but you can use it.</code></pre><pre><code>For Euler/Euler a start with 16 steps max 35 steps (30 steps recommended).</code></pre><pre><code>For DPM start with 19 steps max 35 steps (30 steps recommended).</code></pre>Resolution:<pre><code>Works with all SDXL resolutions, i personally use 1024/1152x1536 and sometimes 1280x1536</code></pre>Recommended base prompts for 2.5D+/3D A+/Simple:Positive:<pre><code>masterpiece, best quality, excellent, high Resolution, aesthetic, bloom</code></pre>Negative:<pre><code>worst quality, bad quality, low quality, lowres, anatomical nonsense, artistic error, bad anatomy, interlocked fingers, bad feet, poorly drawn, censored, watermark, bad hands, extra digits, fewer digits, bar censor, mosaic censor, conjoined</code></pre>Recommended base prompts for Photorealistic 3D:Positive:You can remove 3D if you want skin texture and more realism.<pre><code>best quality, masterpiece, 3D</code></pre>Negative:You can add realistic, photorealistic if you want less realism, more 3d<pre><code>bad quality, worst quality, bad hands, bad anatomy, bad proportions</code></pre>

Mergestein - 3D/2.5D

mergestein3D25D_3dA.safetensors

Files with SHA256: e5cd1a949f07501f97db8eb61b67de8deb80585595a6e3ad05b9042ec99ebc69