This is a Rouwei v-pred based model with a 2.5D Anime style.I decided to create this model because I noticed that there is currently a serious lack of 2.5D based checkpoints on Rouwei base,It excels especially at depicting dark environments and environment lighting.This model is based on the v-pred version of Rouwei merged with some self-made style loras.<h3 id="update">Update</h3> Version 2.0 This version is now based on Rouwei V-pred 0.8. I also made some changes to the style, but kept it rich and colorful.Sometimes the colors can be oversaturated. I will try to fix it in a future release.You need to use a webui that supports v-prediction:ComfyUIreForgeForge (not recommended but works)<h3 id="recommended-settings:">Recommended settings:</h3>Sampler: Euler A / Euler / Euler NegativeScheduler: SGM Uniform / Simple / Align Your StepsCFG: 3-5Steps: 25-30Resolutions: SDXL resolutions (832x1216px, ...)Highres fix: Latent (nearest-exact), 1.5x scale, 15 highres steps, 0.5 denoiseAdetailer depending on use caseVAE is includedThis model works with a wide range of Euler based samplers, but CFG++ gave me some mixed results. Better use the normal samplers.Quality tagsPositive:<pre><code>HDR, raytracing, pathtracing, volumetric shadows, masterpiece, best quality, newest, absurdres, highres,</code></pre>Negative:<pre><code>worst quality, bad quality, low quality, lowres, anatomical nonsense, artistic error, bad anatomy, censored, bar censor, mosaic censoring, watermark, text, signature,</code></pre>Permission: no commercial use, no selling, use only on Civitai

Files with SHA256: 6805684a93863efdba312ee4193ef5176c67bfb5df3c837ba2746a957804a6f0