<p>This is a model that retains the features of anime characters while incorporating realistic lighting and partial realistic textures. Hope you enjoy it, and feel free to share your creations!</p><p>use LazyPos/Neg can get more stable results.</p><p></p><h2 id="ep_remix"><strong>EP_Remix</strong></h2><p>Based on RealWaifuV3 and Dollamor, it has minor anatomical flaws but great style, so I believe it’s worth publishing.<strong>Note that I used an uncommon CFG scale.</strong></p><p></p><h2 id="ep_realwaifu-v3"><strong>EP_RealWaifu V3</strong></h2><p>(It has been merged using LazyNeg and LazyHand. I recommend using the same negative prompt as I do.)</p><p>Details were sacrificed to achieve more realistic anime characters. Perhaps further fine-tuning could improve it?</p><p></p><h2 id="ep_realwaifu-v2"><strong>EP_REALWaifu V2</strong></h2><p>It is recommended to use DPM 2M SDE KARRAS for more realistic textures.</p><p></p><h2 id="edgepaint_realwaifu"><strong>EdgePaint_REALWaifu</strong></h2><p>A broken model with poor details but great waifus :D</p><p><a target="_blank" rel="ugc" href="https://civarchive.com/models/2061990/sketch-style?modelVersionId=2333344">草稿线条风格/sketch style - v1.0 | Illustrious LoRA | Civitai</a><br />This LoRA is recommended for more anime-style faces. You can check the pinned post to view results.</p><p></p><h2 id="edgepaint_real(need-lazeneg)"><strong>EdgePaint_REAL(need lazeneg)</strong></h2><p>I wanted to create a style that looks realistic but avoids 2.5D faces. I’m glad I succeeded, even though this model still has many issues.</p><p>If you are unsatisfied with generation results, make sure you use the same LazyNeg embeddings from the post.</p><p>Details appear broken in this model because I used DareMerge to reduce weights across every block. Perhaps an upscaler can fix this? I’m not sure.</p><p>If you are using ComfyUI, you can copy the nodes to view merge information. Have fun XD</p><p></p><h2 id="edgepaint(need-lazeneg)"><strong>EdgePaint(need lazeneg)</strong></h2><p>It creates wonderful anime artwork with fine details and clean backgrounds.</p><p>No LoRA is required! Just make sure to use the LazyNeg embedding.</p><p>This is a simple 50/50 MBW merge with this model:<a target="_blank" rel="ugc" href="https://civarchive.com/models/2433576/duelillanireal?modelVersionId=2782332">https://civarchive.com/models/2433576/duelillanireal?modelVersionId=2782332</a></p><p></p><h3 id="v2.0-update-(lazyneg-needed!)">V2.0 Update (lazyneg needed!)</h3><p>All LoRAs from V1.5 have been merged into V2.0. The LoRAs originally used for V1.0 are no longer compatible with this version.</p><p>Note: V2.0 is less realistic than V1.5.</p><p></p><h3 id="for-v1.0-and-v1.5">For V1.0 &amp; V1.5</h3><p>Use the following LoRAs for a more realistic style!</p><p>(FOR V1.0&amp;1.5)</p><p>PornMaster-Pro-noobXL - Skin Texture | IL-V1-Lora | Illustrious LyCORIS | Civitai</p><p>(FOR V1.5 ONLY)</p><p><a target="_blank" rel="ugc" href="https://civarchive.com/models/1105685/dramatic-lighting-slider?modelVersionId=1242203">https://civarchive.com/models/1105685/dramatic-lighting-slider?modelVersionId=1242203</a></p><p><a target="_blank" rel="ugc" href="https://civarchive.com/models/1486904/illustrious-realism-slider?modelVersionId=1681903">https://civarchive.com/models/1486904/illustrious-realism-slider?modelVersionId=1681903</a></p><p><a target="_blank" rel="ugc" href="https://civarchive.com/models/1486921/real-skin-slide">https://civarchive.com/models/1486921/real-skin-slide</a></p><p> </p><h3 id="prompt-tip">Prompt Tip</h3><p>I recommend adding these phrases to the <strong>start or end</strong> of your prompt:</p><p>(realistic, real skin, detailed face), masterwork, masterpiece, best quality, detailed, depth of field, high detail, very aesthetic, 8k, cute</p><h3 id="notes">Notes</h3><p>Facial details may lose clarity in medium shots. Face Restoration (or ADetailer) is recommended in such cases.</p><p>This model has anatomical accuracy limitations.</p>

Duel_ILL_AniReal

<p>This version incorporates some of the LoRAs mentioned in v1.5 with adjusted parameters, and integrates an illustration model to improve the facial expression of characters. Overall, it has slightly less realism but significantly stronger expressive power.</p>

Files with SHA256: 30f66550991405363b2a9f50a9e8078cd490a4fe02ccd8d117b97797e6e012ad