Update 1.5: You fool! That wasn't even my final form! finally cut out the burn effect that hurt some prompts at normal CFG like 7, for what I hope will be once and for all, and reduced the darkness even more. No more random super dark prompts and way fewer super saturated images, while keeping the model's personality intact. Done with MBW with <a target="_blank" rel="ugc" href="https://civarchive.com/models/61086?modelVersionId=105782">Vodka v4.0 </a>and <a target="_blank" rel="ugc" href="https://civarchive.com/models/102231?modelVersionId=110716">Centerflex v1.0</a>. Oh yeah, I also have an RTX 3090 now, so expect some fun training from me in the future.<h3 id="heading-1473">1.5 update quick start info:</h3><ul><li><a target="_blank" rel="ugc" href="https://huggingface.co/JPPhoto/neg-sketch-2">&lt;neg-sketch-2&gt;</a> negative embedding highly recommended for realism and 3D style images (among others). </li><li>When prompting for paintings, I suggest using <code>framed, borders, photo</code> as your negative prompt to get fullscreen images and cut out any weird 3D-like people. </li><li>When prompting for illustrations, I like to use <code>photo, realistic</code> as my negative prompt. </li><li>When prompting for realism, I normally use a negative prompt of <code>&lt;neg-sketch-2&gt;, anime, render, pixar, illustration, sketch</code> at 1.2 weight. </li><li><a rel="ugc" href="https://civarchive.com/models/95391?modelVersionId=101827">Garbage-bin LoHa </a>recommended for some any silliness ;) </li><li>Example v1.5 images were generated using latent upscaling from close to ~512x resolution up a few hundred pixels each side at 0.55-0.65 strength w/ DDIM. This was followed up with an ESRGAN model upscale, then converting the image to latents and using ControlNet Tile in a latent to latent stage at 0.2-0.4 strength w/ DDIM.</li></ul>My personal merge of Stable Diffusion 1.5 custom models using the noise offset to improve contrast and dark images. An inpainting model is provided to make inpainting in the model’s styles and detail easier.I like to think of this model as being like base SD 1.5 hyped up on energy drinks and ADHD. Maybe some artist names aside (I don't use them ever, anyway), it has a lot of 1.5's general-purpose power, but has a much higher baseline detail and quality and can be very responsive, provided you're using appropriate settings.This model is meant to be:<ul><li>Artistic and elegant</li><li>Drop-dead easy to work with</li><li>Good at making cool characters and landscapes</li><li>Not bound or leaning towards any single style</li><li>Killer at digital and conventional art in many aesthetics</li><li>And above all, fun</li></ul>It’s not so great at explicit sexual content and anime*, including anime-based embeddings. There’s a million other models for those if that’s what you’re after. *There is some ability to bring out a neat anime aesthetic when you prompt for 'anime style', which I find to be quite cool to look at, although it can be a bit finicky. If you try to make anime-esque art with this model, do not put 'portrait' in your negative prompt, or use 'close' or 'closeup' in your positive prompt, as those seem to force it into a 3d-like style even if you add more weight on the anime style.I want to also bring attention to whosawhatsis' <a target="_blank" rel="ugc" href="https://civarchive.com/models/10540/verisimilitude">verisimilitude,</a> which is great at readily making wallpaper-quality photorealistic stuff.I also want to shoutout coreco and his <a target="_blank" rel="ugc" href="https://civarchive.com/models/1315/seekart-mega">seek.art MEGA v2</a>, which was responsible for much of the composition of V1.3-V1.4, and is an excellent update to his Mega model.Example images were generated in Invoke AI. This means unless you use Invoke AI, you likely won't be able to recreate my images exactly. Just learn from the prompts and modify the weighting in prompts as needed for the UI you use (if you use the A1111 UI, any (plus sign)+ is equal to one set of parentheses).By downloading, you agree to the creativeml-openrail-m and <a target="_blank" rel="ugc" href="https://huggingface.co/dreamlike-art/dreamlike-diffusion-1.0/blob/main/LICENSE.md">Dreamlike-art</a> licenses.Credits (V1.4 / V1.3.5):<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/3738/roboetics-mix">Roboetic’s Mix</a> – Roboetic</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/1315/seekart-mega">seek.art MEGA v2</a> – coreco</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/1116/rpg">RPG V4</a> – Anashel</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/16371">HeStyle V1.5</a> - krstive</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/8067/movie-diffusion-v12">Movie Diffusion</a> - Dalle2Pictures</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/1265/analog-diffusion">Analog Diffusion</a> and <a target="_blank" rel="ugc" href="https://civarchive.com/models/2292/portrait">Portrait+</a> - wavymulder</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/9388/realscifi">RealSciFi</a> - AIfriend</li><li><a target="_blank" rel="ugc" href="https://huggingface.co/Dunkindont/Foto-Assisted-Diffusion-FAD_V0">Foto-Assisted Diffusion</a> - Dunkindont</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/3164/fantasy-art-style">fantasy-art-style v1.8</a> - kasukanra</li><li><a target="_blank" rel="ugc" href="https://huggingface.co/22h/vintedois-diffusion-v0-2">Vintedois Diffusion</a> - Predogl and piEsposito</li><li><a target="_blank" rel="ugc" href="https://www.crosslabs.org/blog/diffusion-with-offset-noise">noise offset</a> – Nicholas Guttenberg</li></ul>

526Mix V1.5

The Anti-Crispy UpdateThis version is a simple dial-back to cool V1.3 / V1.3.5's jets, to reduce the crunchiness, aggressive darkness, general rigidness, and make it less demanding for low CFG and high step count. This is done by a simple merge-back of SD 1.5 (which had an add-difference merge of 0.70 with the noise offset) into V1.3.5 at 15%, and then an extra 2% noise offset on top. This does affect some things -- depending on the kinds of prompts you write, you might not notice much, or you might notice a lot. Interiors aren't as imaginative and smooth, yet making genuine art is a cleaner experience with less crunchiness. With the model's behavior reined in a bit, some things might be a little worse, but nasty crispiness and constantly getting the same posing and framing should be less of a prevailing issue now.Tips:<ul><li>Keep CFG fairly low (e.g. around 6-7, but 10 is too high, things get extra crispy).</li></ul><ul><li>If your results don't look very clean or have an artifact, you'll probably need to re-run with more steps -- this model can be more demanding than others.</li><li>You do not need to prompt a bunch of detail/quality phrases, or use 'negative embeddings' or massive negative prompts.</li><li>To control detail level for less detail-dense art styles, try putting 'photo' and 'realistic' in your negative prompt.</li><li>If your images are coming out a little dark or contrasted for your liking, try putting 'low light' in your negative prompt, or 'dark' for brighter colors if it's not a lighting issue.</li><li>May occasionally need to put 'paint' in negative prompt when prompting for paintings and you're getting weird extra colors or actual paintings in the scenery.</li></ul>Oddities:<ul><li>Eyes and especially teeth are sometimes a bit off.</li><li>Trying for realism with natural landscapes doesn't normally look great.</li></ul>

Files with SHA256: 155cb58446e823b9e530061d9c268cebabff8dbac2f7cedb208192a1cdf2009a