<h2 id="updates">🔄 Updates</h2><p>12.07.26 - There are better Krea2 models on this site , i made this one for me . The format is INT8 ConvRot  , based on TURBO , so CFG 1 ,  8-11 steps  , euler_a - bong tangent ( at 1024x1440 i get ~4s/it on a RTX3060 12GB , FP8 i get ~6.5s/it ) . It very much likes simple workflows ( all my images have them embedded) , for sampler and scheduler see workflow or image details .</p><p></p><p>10.07.26 - V14 best settings in my opinion : Resolution 832x1216 , euler_a with Bong_Tangent , 11 steps , cfg 1 , clip skip -2 , simple workflow .</p><p>07.06.26 - V8 is in lora form ( bastardlora2)</p><h1 id="06.06.26-managed-to-get-v7-to-generate-on-site-but-it-does-not-work-don't-waste-buzz-with-model-just-use-the-loras"><strong>06.06.26 - managed to get V7 to generate on site but it does not work , don't waste buzz with model , just use the LoRAs</strong></h1><p>14.05.23 : Remade the 7th model , if you get black images with F16 model use the repaired BF16 model .</p><p>02.05.26 : Remade the model, should work now</p><p>01.05.26 <a target="_blank" rel="ugc" href="https://pastebin.com/sKxCUSb6">https://pastebin.com/sKxCUSb6</a> script i used , replace base model , variants , weights , density and lambda with what you want . takes ~10 minutes on 32GB ram . If you encounter problems with the model please leave a comment or an image with prompt.</p><h3 id="22.04.26-:-model-(5.5)-still-not-showing-in-model-page-or-my-profile-:))-after-almost-24h-i-sure-picked-the-best-time-to-upload-lol-.-works-with-lora-tested-some-character-lora-as-well-i-have-no-idea-why-some-work-and-others-are-not-working-...-almost-deleted-this-model-with-all-other-model-tests."><u>22.04.26</u> : Model (5.5) still not showing in model page or my profile :)) after almost 24H , I sure picked the best time to upload lol . Works with lora , tested some character lora as well , i have no idea why some work and others are not working ... almost deleted this model with all other model tests.</h3><h3 id="works-best-with-euler-simple-cfg-1-and-9-steps-.-try-cfg-1.1-or-max-1.5-with-9-to-11-max-steps-.-use-a-clip-set-last-layer-node-and-set-it-at-5-)-.">Works best with euler simple , CFG 1 and 9 steps . Try CFG 1.1 or max 1.5 with 9 to 11 max steps . <strong>Use a clip set last layer node and set it at -5 ;) .</strong></h3><h3 id="15.04.26-:-made-an-inbreed-model-it-manages-better-in-my-opinion-.-still-not-nsfw-enough-:(-no-breasts-nipples-or-puss-control-.-loras-not-tested-with-it-.-leave-a-comment-or-a-image-even-with-errors-to-see-what-i-can-do-about-it-."><u>15.04.26 </u>: Made an inbreed model , it manages better in my opinion . Still not NSFW enough :( no breasts, nipples or puss control . LoRAs not tested with it . leave a comment or a image even with errors to see what i can do about it .</h3><h3 id="30.03.2026-:-for-fast-conversion-use-this-nice-tool-made-a-q8_0.gguf-in-&quot;-info:-completed-in-29.14-seconds-&quot;-....-yeah"><u>30.03.2026</u> : For fast conversion use this nice <a target="_blank" rel="ugc" href="https://github.com/qskousen/ggufy">TOOL </a>, made a Q8_0.gguf in " info: Completed in 29.14 seconds " .... yeah</h3><h3></h3><h3 id="25.03.2026-:-managed-to-modify-the-convert.py-script-from-city96-to-make-the-model-directly-in-q8_gguf-.-gemini-for-the-win-.">25.03.2026 : Managed to modify the convert.py script from city96 to make the model directly in Q8_GGUF . Gemini for the win . <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/e3ba59d3-ad1d-40b2-946a-b320da023ccb/original=true/e3ba59d3-ad1d-40b2-946a-b320da023ccb.jpeg" /></h3><h3 id="23.03.26-modified-this-nodes-so-it-can-accept-negative-weights-and-remade-3.5-with-original-zit-for-better-layer-stability-...i-think-.-merged-with-2-random-models-from-tensor-hub-for-aesthetics-poses-clothes-.-i-have-no-idea-what-it-does-better-than-v3.5-but-i-think-it's-different-."><strong>23.03.26 - Modified </strong><a target="_blank" rel="ugc" href="https://civarchive.com/articles/23746/introducing-z-mania"><strong>this nodes</strong></a><strong> so it can accept negative weights and remade 3.5 with original ZIT for better layer stability ...i think . Merged with 2 random models from tensor hub for aesthetics , poses , clothes . I have no idea what it does better than v3.5 but i think it's different .</strong></h3><p>Usage : same as before - euler with simple , CFG 1 with 9 steps (best working and less horror, original text encoder (it does everything you want as long the model knows it).</p><p>Same as before i can't convert to GGUF , if someone knows and manages to do it he can post the GGUF model as his own (with a little mention of me :D ) ...</p><p>(Sample images are 1152x1536 , euler simple, 9 steps , no aura flow , no clip-skip , no LoRA , first run...the first run is bad for me after comfyui 0.17 update so the 2nd run without changing the prompt is better)</p><p></p><h3 id="05.03.26-file-weights-are-broken-because-of-lora-model-merger-node-and-can't-convert-to-gguf-stuck-with-this-..."><strong>05.03.26 - File weights are broken because of lora / model merger node and can't convert to GGUF // stuck with this ...</strong></h3><h3 id="04.03.26-v3.5"><strong>04.03.26 - V3.5</strong></h3><ul><li><p>Sample images are 1024X1536 , 15 steps , CFG 1.5 ( on a RTX3060 12GB i get ~5-6s/it , Prompt executed in ~100 seconds - depends on how long the prompt is)</p></li><li><p><span style="color:inherit">About this version</span></p><h3 id="greatergreatergreaterremake-of-v3-with-new-custom-nodes-from-capitan01r-(leave-a-star-for-him-and-try-the-new-lora-loader-it's-awesome)-+-some-more-loraslesslessless">&gt;&gt;&gt;Remake of V3 with new custom <a target="_blank" rel="ugc" href="https://github.com/capitan01R/Comfyui-ZiT-Lora-loader">nodes </a>from capitan01R (leave a star for him and try the new lora loader it's awesome) + some more LoRAs&lt;&lt;&lt;</h3></li><li><p>Just like V3 most samplers/schedulers work and it has many changes from one to another , test them , mix them , some have body merging some don't , some work for fantasy some work for "erotic" stuff . Try them , try them with CFG &gt;1 , try them with 10+ steps etc. Will update when i get more interesting LoRAs .</p></li><li><p>Sampler : Euler_A / Euler</p></li><li><p>Scheduler : sgm_uniform / simple</p></li><li><p>Steps : 8 - 15</p></li><li><p>CFG 1 - 1.5</p></li><li><p>No Aura Flow Shift (better text render and no skin correction or whatever that shift does ) ,</p></li><li><p>First generation with these settings is sometimes bad without a good prompt but the 2nd or 3rd without changing the prompt is good .</p></li><li><p>It should work with ZIB and some ZIT LoRAs at low strength (did not test) .</p></li></ul><p></p><h3 id="best-settings-for-me-v3:"><strong>Best settings for me V3:</strong></h3><ul><li><p>Sampler : er_sde / Euler</p></li><li><p>Scheduler : Simple</p></li><li><p>Steps : 8-9</p></li><li><p>CFG 1 (sometimes 1.5)</p></li><li><p>No Aura Flow Shift (better text render and no skin correction or whatever that shift does ) ,</p></li><li><p>Added a CLIP set last layer node and i like to use -3 . Every change in clip skip is wild try from -1 to -10 (I'm too lazy to go past that) .</p></li><li><p>First generation with these settings is sometimes bad without a good prompt but the 2nd or 3rd without changing the prompt is good .</p></li></ul><h3 id="26.02.2026-changed-file">26.02.2026 - changed file</h3><h3 id="z-bastard-v3">Z-Bastard V3</h3><ul><li><p>V3 = V2 (identical model).</p></li><li><p>One version was removed because merging and re-merging produced the same result.</p></li></ul><p><strong>Recommended settings (ComfyUI):</strong></p><ul><li><p><strong>Sampler:</strong> Euler , Euler_A , DDIM , res_multistep etc .</p></li><li><p><strong>Scheduler:</strong> Simple , DDIM_uniform , SGM_Uniform , Beta etc.</p></li><li><p>Try <strong>Aura Flow Shift:</strong> 2.5–3 max</p></li><li><p>Also try without Aura Flow Shift — it behaves like a different model and the text is better.</p></li><li><p>I use different sampler and scheduler almost every time, just load the image you like in comfyUI to see and compare with others .</p></li><li><p>Higher Flow shift tends to make the skin more SDXL like .</p></li></ul><hr /><h2 id="how-to-use-z-bastard-v1">🧠 How to Use Z-Bastard V1</h2><ul><li><p>Use it like any <strong>Z-Image Turbo</strong> style model.</p></li><li><p>ZIB-based LoRAs work well.</p></li></ul><hr /><h2 id="how-to-use-unstablebastard-illustrious-v1-and-v2">🧠 How to Use UnstableBastard Illustrious V1 and V2</h2><h3 id="general-generation">General Generation</h3><ul><li><p><strong>Sampler:</strong> euler_A</p></li><li><p><strong>CFG:</strong> 3–5</p></li><li><p><strong>Steps:</strong> ~35</p></li><li><p>Test different samplers — results can vary nicely.</p></li></ul><hr /><h2 id="high-res-fix-script-(comfyui)">🔍 High-Res Fix script (ComfyUI)</h2><p>Using <strong>Efficiency Nodes Highres-Fix Script</strong>:</p><ul><li><p>Upscale: 1.25</p></li><li><p>Steps: 12</p></li><li><p>Iterations: 1–2 (2 is my favorite , refined from 768x1152 to 1200x1896 in ~120s with my 3060 12GB)</p></li><li><p>Denoise: 0.26-0.36</p></li></ul><hr /><h2 id="face-detailer-(comfyui-impact-pack)">👤 Face Detailer (ComfyUI – Impact Pack)</h2><p>Recommended settings:</p><ul><li><p>Steps: 8</p></li><li><p>Denoise: 0.26-0.36</p></li><li><p>Lower chance of male faces becoming feminine</p></li><li><p>Works with upside-down faces</p></li><li><p>UltralyticsDetector Face <a target="_blank" rel="ugc" href="https://huggingface.co/luxdelux7/ForbiddenVision_Models/tree/main">model</a></p></li></ul><p>If using ComfyUI, just load one of my images (workflow is embedded) and run it.</p><hr /><h2 id="realistic-images">🎨 Realistic Images</h2><p>For realistic outputs from ill with v1 and v2 i use :</p><ul><li><p><strong>CyberRealistic / CyberIllustrious</strong></p></li><li><p>Highres-Fix Script enabled with one of the checkpoints above</p></li></ul><p>You can use any checkpoint with this setup.</p><hr /><h2 id="artist-prompt-examples">🎭 Artist Prompt Examples</h2><p>You can improve style by adding artist tags like:</p><pre><code>(artist:wamudraws:1)
(artist:slugbox:1)
(artist:zankuro:1)
(artist:mazjojo:1)
(artist:kotteri:1)
(artist:pigeon666:1)
(artist:makoto-shinkai:1)
(artist:zawar379:1)
(artist:yoneyama mai:1)</code></pre><hr /><h2 id="weaknesses">⚠️ Weaknesses</h2><ul><li><p>Not great with complex backgrounds</p></li><li><p>Struggles with fine detail-heavy scenes</p></li></ul><hr /><h2 id="lora-strength-guide">🔗 LoRA Strength Guide</h2><ul><li><p>1–2 LoRAs → <strong>0.5–1.0 strength</strong></p></li><li><p>Multiple LoRAs → <strong>0.2–0.5 strength</strong></p></li></ul><p>(Adjust as needed.)</p><hr /><h2 id="prompts">🧾 Prompts</h2><h3 id="positive-prompt-(my-go-to)">Positive Prompt (My Go-To)</h3><pre><code>embedding:lazypos,
(masterpiece, best quality, amazing quality, ultra-HD, high detail, very awa,
newest, very aesthetic, ultra-detailed, absurdres, 8k)</code></pre><p>(Uses 1 embedding.)</p><hr /><h3 id="negative-prompt">Negative Prompt</h3><pre><code>embedding:lazyhand,
embedding:lazyreal,
Bad Quality, worst Quality, low Quality, photorealistic</code></pre><p>Note:</p><ul><li><p>You <em>can</em> use <code>embedding:lazyreal</code> in the <strong>positive prompt</strong> if you want a more realistic feel.</p></li></ul><p></p>

UnstableBastard

<p>Made this one to work with character LoRAs trained on this site . Below are the settings i used ..shit's scarry good (made one of myself and it's too accurate) With 45-50 images (i didn't use any good ones or 1024X1024 ) At Steps i always make sure every image is seen 65+ times , @Checkpoints i put 15 for good measure . For image generation i use the last 2-3 at low strength (best one at 0.6 and the rest at 0.3) . All images were captioned (i did not use tags like the site suggests ) on site. </p><pre><code>{
  "lr": 0.0001,
  "steps": 3400,
  "engine": "ai-toolkit",
  "epochs": 15,
  "batchSize": 1,
  "ecosystem": "krea2",
  "keepTokens": 0,
  "networkDim": 32,
  "resolution": 1024,
  "lrScheduler": "constant",
  "minSnrGamma": null,
  "noiseOffset": null,
  "networkAlpha": 32,
  "optimizerType": "adamw8bit",
  "shuffleTokens": false,
  "textEncoderLr": null,
  "flipAugmentation": false,
  "trainTextEncoder": false
}</code></pre>

Files with SHA256: 702376000b246fa170d0d03ec56a8b12aff162a30924b3d1a59c49d0160a93bf