Cacoethes

***MY GUIDE TO MAKING HIGH QUALITY IMAGES WITH SDXL FOR FREE*** 04/09/24 Update with new recommended models! You've probably seen some of the models in PixAI's directory with *FREE* or *ZERO CREDITS* in their title. If you haven't, they are easy enough to find. The problem I've found with those models is that they are mostly tailored to animé style images. If that's all you aspire to, then go grab them, and knock yourself out! I won't judge you - honest! But... if you like to create more sophisticated imagery - with a view to realism - then read on... These models all produce excellent results when using Euler A sampler with just 7 steps (so a batch of 4 images will cost zero credits, unless you choose the square format). The list is not exhaustive, and I will update it with new models which live up to my aesthetic criteria as and when I discover them... - Dream Art Fusion - https://pixai.art/model/1772034101963669694 <<< My new favourite model!!! - Quantum Fox XL - https://pixai.art/model/1761047935347656838 <<< arguably best-in-class for photorealism - AlbedoBase XL - https://pixai.art/model/1718826133614653525 <<< excellent all-round - Atomix XL - https://pixai.art/model/1724983329338593206 - Divine Diffusion by Dever (DDD) [SDXL] - https://pixai.art/model/1773720349367014162 - DreamShaper XL - https://pixai.art/model/1718828681490755821 <<< Recent addition - FULLY_REAL_XL v5.1 Human - https://pixai.art/model/1698480930363325045 <<< Recent addition - Halcyon SDXL - Photorealism - https://pixai.art/model/1778790045054190086 <<< Recent addition - Imaginarium v1.0 Turbo - https://pixai.art/model/1698873791500814760 <<< solid, adds artistic flair - Jib Mix Realistic XL - https://pixai.art/model/1773409362483691835 <<< NEWLY ADDED - Juggernaut XL - https://pixai.art/model/1718821551370739798 - Moxie Diffusion XL - https://pixai.art/model/1771188981655114073 <<< another strong all-rounder - MySLR - https://pixai.art/model/1772035109527765482 <<< From the author of Dream Art Fusion! - RealCartoon-XL - https://pixai.art/model/1684765292480180319 <<< solid all-rounder - RMSDXL - Hybrid Turbo XL [base_model] Aries 🎨 - https://pixai.art/model/1780620382186421668 <<< NEWLY ADDED - RobotDream - https://pixai.art/model/1763337019752993181 <<< Recent addition - SDXRealisticA - https://pixai.art/model/1773423431924072619 - Turbo Diffusion XL v1.32 - https://pixai.art/model/1722904403644014341 <<< NEWLY ADDED - Valhalla [SDXL Lightning] v1.0 - https://pixai.art/model/1784125314812078744 <<< NEWLY ADDED - Zaxious_XL Wild_Child V3.0 - https://pixai.art/model/1774430998188545313 <<< very strong all-round, but likes to surprise you! *NOTE* - some of these models will automatically overwrite your sampler settings and negative prompts when you select them, but you can (and SHOULD) adjust them manually to Euler A, 7 steps. Always remember to check that the credit cost is ZERO before you click on the generate button And since you're going to be using Euler A for your sampler, I highly recommend this LoRA - https://pixai.art/model/1687631572599847061 Set the weight around 1.0, and it's like you just doubled your steps count! I never even think of starting a task without this LoRA This will still leave you with 2 open LoRA slots (or more if you're a PixAI Member), so you could maybe use a style, and a character LoRA, but if you REALLY want to take your art to the next level, then try adding one or both of these enhancement LoRAs: - Detail Tweaker XL - https://pixai.art/model/1668692278399247652 <<< (strength 0.7 if used in conjunction) - ModelBoosterXL = https://pixai.art/model/1725049259066326012 <<< (strength 0.8 if used in conjunction) I keep my CFG low (under 2, but I find 1.1-1.5 works best for the first 2 models, and 1.3-1.7 is best with RealCartoon), but depending on the LoRAs you are using, and the style you're trying to achieve, you might need to go a little bit higher. When using additional enhancement LoRAs, you may need to increase your CFG a bit. If you use all 3 together (as I currently am), a good range for your CFG is between 1.7 and 3.0. But by all means experiment with weights and LoRAs until you find something that is right for YOU. If you are using my recommended sampler settings, you can experiment without having to worry about running out of credits Waiting times can be epic, but the results are usually worth it ;) Apart from my first few images when I had no idea what I was doing, everything you will find in my gallery was generated without spending a single credit. None of them have used a reference image, in-painting, or upscaling. I have over 4 million credits stashed, and I've never purchased any! STOP SPENDING YOUR CREDITS ON IMAGE GENERATION, AND USE THEM INSTEAD TO SKIP WAITING TIMES, OR FOR MAKING YOUR OWN LoRAs! For prompts, you can use ChatGPT. My requests usually start with something like "generate a stable diffusion prompt to create a photorealistic image in the style of [insert name of famous photographer], depicting...", and then some details about the subject matter. Doesn't have to be an essay - sometimes less is more! Copy & paste the response, edit it as necessary, and add your usual stuff like (((masterpiece))). If you don't like ChatGPT's approach to prompt writing, then why not check out this free* amazing guide to writing prompts for photorealistic images which I found online? (*you can name your own price to download it as a .pdf, just select $0 if you don't want to pay for it. You can always re-download it again and buy him a coffee if you found it useful) - https://promptgeek.gumroad.com/l/photoreal That's it! Now go and make some amazing art! Just remember, this guide is only a guide. Make art that makes *YOU* happy If you found this guide useful, throw me a few likes, or just drop a comment on one of my images ^-^ "Remember what the dormouse said... Feed your head! Feed your head!" *** CREDITS & SHOUTOUTS *** I've written this guide entirely by myself (so ALL errors are my own), but I didn't find all of the resources in a vacuum, so let me say THANK YOU to the following artists - inevavae 🍎🐍 - https://pixai.art/@inevavae/artworks <<< for contributing most of the models in my list - Danklaw - https://pixai.art/@danklaw/artworks <<< for contributing the additional LoRAs They both create incredible artworks, in many different styles, and you really ought to go check out their galleries. If you have found yourself stuck in a rut of creating generic animé pics of your favourite manga or MMORPG waifu, then open your eyes to what AI art is REALLY capable of, and be inspired! Lastly, I'd like to give credit to the people who actually MADE the Models and LoRAs in my guide. Maybe one day I will get around to naming them all, but today I am too busy MAKING HIGH QUALITY IMAGES WITH SDXL FOR FREE!!!