The EASIEST way to generate AI Art on your PC for FREE!

stable diffusion the air generator with some modules that can create some crazy accurate to your prompt results has fully launched publicly available with open source stuff and a lot of different tools have come out and we're going to be navigating quite a few of them here on the analog dreams youtube channel but i wanted to start out with the easiest most accessible form and in this video i'm showing you how you can generate this the best pizza in the world or david harbour as thanos with basically just a double click on windows here with as little setup as possible this is the easiest way to run stable diffusion locally on your machine without any issues i am super stoked i'm addie welcome back to analog dreams the channel dedicated towards empowering your art and creativity and walking you through some of the crazy glitch art and ai you know art generator tools and eventually we'll focus on like affinity software and stuff like that but before we get to that this is stable diffusion g-risk gui this is the project that we're focusing on here today this is available on itch.io and you do need a nvidia graphics card in order to run this as it leverages the cuda rendering engine that only nvidia uses uh ideally someone will make some of the stuff that runs on amd or intel in the future but a lot of this stuff really kind of works best with nvidia cards anyway so unfortunately you do need that but otherwise you don't need to do much of anything it's kind of dubious how they're doing this but you you just go to download download the three gigabyte dot rar file save it to your downloads open it up in your favorite archive manager or maybe windows iu 7-zip it's free open source linked below and extract the files they'll extract to their own folder and you just scroll down and run the stable diffusion g-risk gui dot exe that is it and you are running it'll pop up a little command prompt to run the script and you get this little gui program right here the ui is mostly fairly straightforward but there are some you know stable diffusion specific things we will take a look at you can choose an image model if you import your own into here by default it only has the one i'm not going to mess with additional ones but eventually you will be able to be more modular with it you have your text prompt location that you enter your text into you can choose your output folder by default it just goes to a results folder in the folder that you just extracted here it's fine to leave it but you can you know make a folder on your desktop or something if you'd prefer that then you have your steps your vscale your samples per prompt which is currently broken so it won't do anything if you change that it'll only ever generate one image you can use seeds which we'll talk about in a moment and then your output resolution the more output resolution you use uses more vram i'm on an nvidia rtx 3090 which has 24 gigs and i can't exactly go that big especially when i have all my recording stuff open at the moment but we're going to come up with a prompt here we're going to say a computer's dreams and imaginations so you have two main options to configure here you have steps which are basically how long it will take to create the image and to create a more detailed image the app actually recommends about 150 steps or less some of the people on my discord server say they have better with closer to like 30 to 50. then you have v scale which is how much it's going to adhere to your specific prompt and if you set something absurdly high like 50 it's going to basically like deep fry everything and just maybe not get the result that you want 7.5 is the default between 5 and 7 seems to be the best result so far from the people that we've tested with i haven't done a whole lot of thorough testing but you can experiment from there but for steps we're going to do like 100. so we've got a computer's dreams and imaginations 100 steps which again is like quality how long it takes to make the image it does not use more vram for more steps and then you have your output resolution most lower end graphics cards with like eight or less gigabytes of vram shouldn't really experiment with this too much but we can do like 1920 by 1920. now i am using most of my vram already here i'm using 16.2 for all my obs scenes and things like that uh but that shouldn't be a huge problem we're going to go ahead and fire a render and see if i crash everything a few moments later out of memory yep yeah we ran out of memory because of obs going so i'm going to pause all my recordings and we're going to come back when it can actually make the image but when you do generate an image it will give you a png file in the resolution that you set i could probably set this back to 512. we're going to fire that render off it'll give you the png of the image that you set as well as a text file with all of the configuration the text prompt that you gave it the folder it's stored in the resolution and all the options just so you can reference that whenever you need to for future results so we have a render go in here at 5 12 by 5 12. you can see here it's going uh it would be the steps that it's on so it's on step 44 how long it's been going and how many iterations or steps per second that it's operating at which is really good information you can kind of start to binge market a little bit too we're at step 63 and we told it to do 100 so we're almost done and we're done and it gave us something with a bunch of text for some reason the rest of the image would actually be kind of cool then you can iterate from there the amazing thing about this and why i have been advocating for this and in my big video about ai art tools i mentioned building my own dedicated ai rig to run things on i'll have tutorials for running stable diffusion and some of the modules and stuff on your own thing if you want to do a hardcore linux server setup this is just supposed to be for newbies running windows not meant condescendingly but just you know this is the entry level version this costs you nothing you are running it locally on your machine you are in control of what's going on there's no filters or restrictions there's no credits that you have to renew you're just running it locally you just need the power and the patience and you can fire off like a bunch of these like for example i can say computer streams and imagination no text a computer dreaming i can do pizza sauce and then what do we want what do we want what do we want we want david harbor we want a realistic oh beautiful portrait of david harbour as mickey mouse and it will fire off all of those as separate renders so if you want to like for example spend your day while you're at work or something taking notes in your own little notepad or sketchbook or whatever kind of coming up with the ideas of the prompts that you want to generate and then before you go to bed at night just you know add them to a queue fire it all off you get to wake up and have just like a million images generated at no cost just waiting while you were asleep it's freaking awesome we're going to fire off these renters it's gonna take a hot minute cause we're doing four all right it has finished our queue and i will say right up front that mid journey does a lot better of a job than a lot of these other services at more abstract just random thoughts prompts versus stable fusion and dolly do a lot better the more specific and like very detailed you get but here is our results we have a computer's dreams and imaginations with no text a computer dreaming pizza sauce and david harbor as mickey mouse i guess the broad shoulders part is supposed to be the david harper part of that mickey mouse but otherwise that's just a mickey mouse pizza sauce is a pizza you get the idea though it's a really cool tool you got to play with it and finagle it a little bit and if we if you do want to come over to the hardcore side of running the more linux focused python tools uh once we get that set up you can add in a bunch of modules and things get absolutely wild once you go that way hope you enjoyed this little intro or teaser or what have you a lot more cool stuff coming let me know what you've been making with stable diffusion in the comments below or over on our ai experiments channel on discord discord ggsheeplesbox remember to be kind rewind thank you for your support here on analog dreams

Channel: analog_dreams

Views: 50,079

Keywords: stable diffusion, run stable diffusion locally, stable diffusion easy, stable diffusion made easy, stable diffusion gui, grisk gui, stable diffusion tutorial, how to run stable diffusion windows, stable diffusion guide, stable diffusion setup, stable diffusion free, stable diffusion tips and tricks, how to make ai art, how to generate ai art free, free ai art generator, best ai art generator, ai art tutorial, ai art for newbies, analog dreams, tutorial, art guides, art tips

Id: tN7gGk_h93M

Length: 8min 28sec (508 seconds)

Published: Fri Sep 02 2022