<h1 id="heading-65">Welcome to Taureal Mix!</h1>--If you love my work you can buy me a coffee : <a target="_blank" rel="ugc" href="http://ko-fi.com/tauron">ko-fi.com/tauron</a> you dont have to, but if you do, it helps me stay awake until midnight (which is the only time im able to work on these stuff lol) -- or.. a like or a review wont hurt or takes you more than 1 mins, while it takes me hours, days or even week to make this - it helps the thread alive and afloat.--**Update Changelog can be seen at model version details.**NOTE : Alot of sample images has been removed by CivitAi even though there is no TOS violation.--<h3 id="heading-12371">Suitable to those who needs:</h3><ol><li>All round realistic model with versatile styles.</li><li>Low memory cost.</li><li>Accurate txt2img results with short/long prompt.</li><li>Low model bias = more accurate lycoris/lora representation.</li><li>Better integration with my previous / future lora.</li></ol><h3 id="heading-12372">Disclaimers:</h3><ol><li>tested mostly for txt2img, for other functions use at your own discretion.</li><li>tested specifically for my previous or future lora. loras from other creators may work or may not work well with my model.</li><li>a pruned model is not suitable to be merged. choose the v3.1 or v2.1 instead.</li></ol><h3 id="heading-12374">Prerequisite &amp; Resources</h3>These are the resources i use for my standard workflow :<ol><li>ADetailer<a target="_blank" rel="ugc" href="https://github.com/Bing-su/adetailer">GitHub - Bing-su/adetailer: Auto detecting, masking and inpainting with detection model.</a></li><li>Negative Embeddings (Use 1 only)Use when needed only, v3.7 pruned without negative ti is observed to produce significantly better result compared to v2.1 / v3.1<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/4629/deep-negative-v1x">Deep Negative V1.x - V1 75T | Stable Diffusion Textual Inversion | Civitai</a>accurate prompt such as race, skin tone but smoothen the skin texture and has face bias (change result face), also sometimes misinterpret prompt position occassionally.</li></ul><ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/7808/easynegative">EasyNegative - EasyNegative | Stable Diffusion Textual Inversion | Civitai</a>most accurate amongst other negative ti -- prompt such as race, skin tone, face, and texture is very accurate (90%), but sometimes misinterpret prompt position (ex. cum is supposed to be on body, but is on face instead)</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/17083?modelVersionId=20171">bad-picture negative embedding for ChilloutMix - 32 Vector Version | Stable Diffusion Textual Inversion | Civitai</a>most detailed amongst other negative ti -- this embedding (32v or 75v) pushes skin texture and facial detail to the top. however, it changes skin tone, color, and face significantly producing inaccurate result if you are trying to use lyco.</li></ul>*DO NOT COMBINE NEGATIVE - TI WITH SAME FUNCTIONS IN THE V3.7 MODEL*</li><li>Facial Enhancer Embeddings (optional)<a target="_blank" rel="ugc" href="https://huggingface.co/yesyeahvh/ulzzang-6500">yesyeahvh/ulzzang-6500 · Hugging Face</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/4514/pure-eros-face">Pure Eros Face - PureErosFace_v1 | Stable Diffusion Textual Inversion | Civitai</a></li></ol>Credits to all creators above that helps the model achieve its sample images quality above.<h3 id="heading-12375">How To prompt this mix</h3>You can follow your own prompting method, this guide only serves as a convention when you are having trouble producing similar results with my lora.<ol><li>Style —&gt; realistic , realistic details, detailed for realism / anime for anime / masterpiece , best illustration for conceptual art.</li><li>Subject —&gt; describe your subject can be anything. 1girl, solo, cute, beautiful, slender, etc. if you use my character lora this should be where you put trigger words of the character</li><li>Attire —&gt; describe the outfit of your subject. If you use my attire lora this is where you should put trigger words</li><li>Background / Context —&gt; describe the context or the scene</li><li>Textual Inversion &amp; LoRa brackets</li></ol><h3 id="heading-12376">How to maximize this mix?</h3>Sampler &amp; Steps<ul><li>SDE Karras Sampler = 16 / 8 Denoise @ 0.32</li><li>DDIM Sampler = 16 / 8 Denoise @0.32 (Observed Fast like 2M, but with SDE Quality, DDIM is very good for concise prompt, but is weak at translating complex prompt)</li><li><s>Using negative embeddings is a must</s><s> to create sample image quality. </s> Negative Ti is no longer needed to achieve sample image quality</li></ul>CFG Scale<ul><li>CFG Scale (creative) 4 ----------- 10 (accurate) -- 12 (Max)</li></ul>Resolution &amp; Upscaling<ul><li>Resolution: 512 x 768,</li><li>Upscaling = 4x_Ultrasharp <a target="_blank" rel="ugc" href="https://huggingface.co/lokCX/4x-Ultrasharp">lokCX/4x-Ultrasharp · Hugging Face</a></li><li>To get even better results: Use Lightroom to edit your result.</li></ul>Utilizing ADetailer<ul><li>Face = Noise 0.3</li><li>Hands = Noise 0.25-0.4</li><li>Body = Noise 0.3-0.5</li></ul>The rule of thumb is that there are no constant equation, you need to find one your own that suits what you are aiming, and sometimes you change these value on the fly depending on your goal. its better to know what these parameters do to your results rather than copying values from people's post.--As an afterwords, i hope you enjoy using this model, a like or a review definitely helps, or if you find it distasteful, you don't have to use it. there are many models out there that produces similar or superior results.

Taureal Mix

More model optimization: <ol><li>Lower dependencies on negative Ti (now only use ng_deepnegative_v1_75t + badhandv4</li><li>Better &amp; more accurate low CFG Scale (5 ~ 10)</li><li>Wider Dynamic CFG Threshold (10+ needs DT)</li><li>Additional Quality bump on skin with fantasy / sci-fi prompts</li><li>Better Wet &amp; Water effect (at SDE sampler)</li><li>Very Nice Quality with SDE Karras with Low Sampling Steps (16 / 8) with 4x-Ultrasharp can match v3.1 (50 / 25) with ERSGAN_4x+</li><li>Observed very high quality output with DDIM with (64/16) steps. DDIM is faster to calculate and produce more creative and detailed result than 2M Karras and is similar to SDE Karras result. However, it can be very tricky and hard to control the result at the moment - Use this to find a good seed.</li><li>Hotfix : VAE is baked in</li></ol>

Files with SHA256: 537f5ddce6845ed552208ea90a71a0286956104115559f75c3c49f25f96e0a2f