HOW TO CREATE PHOTOREALISTIC AI IMAGES | Stable Diffusion

hello everybody my name is Binks and welcome back to another video today we're going to be talking about stable diffusion and a photorealistic workflow that I've been experimenting with over the past couple of days I think you guys are going to love the results um this video is going to be less of kind of a tutorial I will be showing you all the settings like always and down in the comment section will be a copy and paste for all the for the prompt and the negative prompt um but outside of that we're just kind of going to experiment a little bit and I'll show you a couple of things that you can do if you're ever confused about any of the things that I'm doing in this video I have a link up in the top right for the playlist of all of my stable diffusion videos so go feel free to check that out subscribe comment like for more and let's get right into it so one of the first things I've been changing up over the past couple of days with my stable diffusion prompts is I've gone from less of a keyword approach and more towards something that would be closer to a large language model I've been messing a lot with gbt3 and chat GPT from open Ai and that's been a lot of fun so similar to how those sort of prompts and text completion works I wanted to take something that was more of an English structured sentence into stable diffusion and see how it did and I've actually been getting some really great results from that I can also go ahead and paste in a negative prompt here so I'll just go ahead and throw that in there real quick we're going to use the DPM plus plus SD Kara sampler because that is my favorite as everybody knows I'm always turn our batch count up to two if you know you know width is going to be 768 by 768 just a little bit higher res than what is considered normal a convex scale of seven is fine for me I'm going to make sure to check restore faces and go ahead and generate foreign s are finished and before we even say anything else about them they're obviously stunning and you might be wondering Banks can I just load up whatever version of stable diffusion 1.4 1.5 2.1 and do this no uh sadly not you would have to download this model that I'm using today which is realistic Vision version 1.2 link will be down in the description it's from civet AI um a couple of words of warning though just be careful there is NSFW content on the site and if that's not what you're looking for then it can be kind of in your face and you can you can disable all of it but like I said it's it's a little it's a little forward big shout out to sg16122 you're the best this uh 1.1 and 1.2 models have both been great and there's a lot of inspiration here on this site if you guys are interested it's a 3.8 gig download not too too shabby considering some of the bigger models are closer to 7 to 10 gigabytes so for what it does it is really great there's a couple the the only con that I found so far is that it likes to generate similar faces um even particularly when you do an image to image with too high of a denoising strength if you give it too much Freedom it really likes to drift away from the original subject and kind of head towards one of its training images I think that could be solved in future updates so just keep an eye out for this page and um if there's ever an update I'll probably do a video on it as well okay so let's get back into stable diffusion here as you can see these results are absolutely astonishing they're great um and right here just off the bat we could send any of these over to an upscaler and you know go from there but what I want to do is kind of learn a little bit more about realistic vision and the way that it ticks so we're going to start modifying these prompts a little bit the negative prop seems to have done a straw rather well I don't see any really obvious there's maybe a weird thing with the eyes here but that might be a perspective shift thing um honestly these images are great so I'm just going to change this prompt up a bit okay so I've just kind of Switched The Prompt around um a man brown haired man with striking features short beards scars on his face graph expression just kind of a complete different thing and we're gonna let this run and I'll kind of show you how versatile this this model is foreign [Music] and there we go we've generated two more images this time of a male it seems like we've got either once it's trying to do a tattoo or maybe hair uh but you know cropping up on that neckline will probably be better and then the left variant is also very nice um not exactly what I was looking for though I think I would uh delete this last line and let that rip again and there we go we got two more results that I think are great I've actually been using AI a lot for my world building I do like hobbyist World building and I'm designing a medieval fantasy world for a game that I'm working on as well and this AI has been great for inspiration so I just kind of wanted to show you guys a couple of things you can do to keep having fun with AI don't don't get discouraged with stable diffusion it does take a little bit more time to get used to and kind of understand but I'm going to keep the content coming if you haven't seen all my other videos on it please take a look at those I've had a lot of people say that they've been extremely useful for them and if you guys have any questions at all make sure you leave a comment down below like And subscribe and I'll see you guys in the next one thanks for watching

Channel: Binks

Views: 152,387

Keywords: ai, stable diffusion, midjourney, how to use ai, ai tutorial, ai art, ai art tutorial, protogen, photorealism ai, free ai art, chatgpt, machine learning art, how to make nsfw art, ai nsfw, twitter art, civitai, lexica art, stable diffusion better models, better ai, improve ai art

Id: s4YIUaIRyF8

Length: 6min 1sec (361 seconds)

Published: Fri Jan 27 2023