Illuminati Model with Noise Offset & Weekly AI Art Challenge

Key Takeaways

hey hey so there was this uh joke writing contest and I something 10 pounds to see if any of them would win sadly no pun in 10 did while we're talking about contests I have a weekly challenge on my Discord and I figured it would be a great video to actually do that challenge you can watch me create the submission for the challenge I'm actually gonna do that with a new model that I've been checking out lately which is the Illuminati model so this is going to be a twofer we're going to check out the Illuminati model and we're gonna see a workflow where I make the submission for the weekly art challenge which for this week is dystopian and if you want to take part in this challenge just hop on Discord and check the challenges channel so here's Illuminati diffusion you can see some example images and as you can see most of them are fairly dark and it's using noise offset which uh is a new feature that can keep you know the contrast with the dark or the light inside the image in in regular stable Fusion if you have a light background it tries to go for a darker foreground and vice versa and here they have offset that in well it's called noise offset so just download this put that inside your stable Fusion web UI models stable Fusion drop that right in there then you're going to download the textual inversions which is these three files and then you're gonna put them inside a stable Fusion web UI embeddings and the way you load embeddings is you just type their name so we're going to type in real fixer and and fixer then I'm going to take the prompt that actually would be saw in I think it was this image so you can see all the problems here almost all the settings so I took this prompt I just remember that this is a 2.1 models you can't use control net with this as of recording this video maybe you can in a bit and this is a 768 model and we're actually going to make this a little higher so 768 10 24. so we have dystopian theme for this week's challenge so we left a lot of the prompt but we changed the movement to a photo of a man and dog in dystopian Fallout epic scenery Sunset and then we're gonna run a lot of batches and what I'm doing now is I don't have a specific image in mind that I want to create or a specific composition I have a specific idea of what I want but not what it should look like so I'm just getting a lot of images out to get a feel of what we're doing and then we're gonna improve on them later and that's a big dog I think actually I am going to go back into I want to flip this to a horizontal one instead because I feel that we want to see more of the scenery more of the landscape and I'm not really getting that and as we're rendering this you can see that all the images are fairly dark and that's kind of consistent with this Illuminati model everything is a dark and gloomy even though we have some neon signs and rainbows shiny reflective whatever inside of this so let's see we have it we have 10 images and I'm kind of feeling this is kind of cool we have the man looking back and the dog looking straight at us and now we have a big neon sign here so that's nice I'll also like this one here I was gonna say I like having the the water because of the reflection we have water in almost all of them but for now I think we're gonna try and keep working with this one and we'll see if it works or if we need to go back and pick something else I think this Vibe is pretty dramatic as well the dog looking at the man here anyway let's see which one was it here yeah that's cool okay we're gonna send this to image to image and then I'm going to use my usual workflow which is uh doing this a little bigger I'm not going to be super consistent with the size here now or the aspect ratio rather I'm just gonna go up to full HD I think we're gonna do four at a time no I think we're actually gonna do two at a time because last time I did this I had problems with the recording and I'm gonna generate all right so it's a little slower than usual because of OBS but uh I guess that's okay while that's working we can check out the current submissions in the challenge right now so this is my Discord and uh here are all the dystopian challenge submissions as of right now I like this one this is great this is nice too very last of us this is nice as well this feels very I want to say realistic but cinematic realistic it feels like a movie very nice some kind of a night this is very nice and with the light here as well he probably used the Illuminati model they've been playing with it for a while now so let's go back and see what we found here I think this is kind of cool but I don't know what's going on here kind of turn into a jacket and with two dogs in it here's an extra dog oh and this is all crazy we have two people inside three heads and this one's part dog and two dogs here and one dog here so while the images are cool they're kind of broken so we're gonna lower the denoising strength here and we're gonna run two new ones but first let me tell you about the sponsor of today's video which is opencv the largest computer vision library in the world wow they have launched a Kickstarter with courses in AI they have fundamentals of generative AI they got Air Generation using Dolly learn AI generation using mid-journey guides in dream Studio how to generate infinite images with stable Fusion wave UI a deep dive into prompt engineering with stable diffusion control editing of images like object removal used in painting and inserting objects using masks examples like paint on an infinite canvas without painting and virtual trials where you can change the dress ties accessories anything even with virtual hairstyles create personalized AI generator avatars with dream Booth or you can learn about trading textual inversions learn how to create animations and videos from images prices start from 29 up to their supermaster bundle at 9.97 their super master bundle includes all the courses opencv has ever offered click the link in the video description and check out their Kickstarter I think we probably squished the aspect ratio too much so we might need to go back and fix that you can see it already so I'm gonna interrupt this actually I mean it worked with the high denoising strength but if you go back it's um it's squishing it so what did we have we had 768 right 10 24 768 2048 15 36 right maths this might be a little big for my computers I'm gonna do one batch for now out of memory too bad so let's just go up to 1500 1125 which is this the same aspect ratio that we had before I think was four by three I think we should be able to do two now and we have success they're fairly similar both of them because we have such a low denoising string I'll just pick um I'll just pick this one because we had some more bokeh and I think this is kind of cool to be honest all right cool let's send this to him painting and I'm gonna try and change the sign here now it's going to mess up the water for me I'm afraid but it is what it is so we have neon and sign we're gonna impaint only masked here I'm gonna up the denoising so we can get some changes into this and uh well let's see what happens so we have this as soon it falls here which I can assume is some kind of advertising for uh sports drink or whatever now our water here is uh similar to what was before I mean it's not perfect you can see that it's not exactly the same um letters but uh we're gonna be happy with that for now we got some more detail in the sign here and I like that I think the exit Transit here whatever it says um I think we'll keep that for now we're going to give some more detail to the face here so we're in painting that we're gonna remove this we're gonna say man head we're gonna lower the denoising since we just want more detail into this we have original and we have only masked here so that will draw just this instead of the whole image which will give us more detail so yeah that's fairly cool we're gonna do the same thing with a dog here I'm going to take all of this dog actually gonna move neon signs we don't need that here now it won't probably won't put neon signs in but anyway and after denoising just a little bit I don't know a couple more steps and don't forget to drag your image to the left side here I think we well we got some of the detail but we lost the harness here but I mean I guess that's fine it's more of a dystopian dog anyway however the purple eyes are kind of weird but um that'll do for now so we're gonna take this and we're gonna send it to photo P just for some finishing touches and then we're gonna have a look at uh some other stuff in this Illuminati model you see behind my face here or next to my face it's adjustment layers and you want the curves you want to get some dark down here you want you can see if you move this up you get more light and remove this down you get dark and down to the left those are your darks and here are your lights so see if you erase this your darks gets brighter so I usually want to get the darker parts just a little darker I want to raise some of the mid here and some of the mid up here it all depends on your image as you can see here if you move it too much up here he's graying out the image it's going to keep it where you get a nice contrast down here but you want this high so you get you know the nice lights you can see it behind this face and on the background so before after before after before after now this is um a lot of contrast but I think for this kind of style with the upright lights here it's fine then we're going to add Vibrance we're gonna increase this there we go something like that and then as usual Ctrl shift alt e going to filter sharpen unsharp mask not as high as this 75 at about 2 pixels I think actually for this image it's gonna do enough with one pixel and there we have it this is going to be our submission for this week's dystopian Challenge and let's go back into stable fusion and see if we can create something else go back here and we're going to try with our favorite style keyless the oil you can find this in my Discord in the resources Channel let's try and do astronauts flying in a black hole and we're just gonna do a square for now ah 768 768. do four images I mean it's kind of coolish but um very far from the Illuminati model in general but it's a very artistic illustrative style I like that actually let's go back and check some of the prompts here okay let's try this one something I noticed now the cfk scale is very low in all of these and a fairly short prompts so let's try that I'm gonna add a glossy rainy RTX octane unreal I'm gonna change this into astronaut on ancient city street okay I think uh in regard to prompting this has been trained in a different kind of way so look at the prompts that are available for this specific model and use that I think this is pretty cool this is a more of a was not an ancient city but it's more of an older style City and it um it's like an astronauts just traveling through time and popping up here surprised let's put him in uh astronaut in in ancient Egypt yeah I kind of Simon all of them this kind of turned into a mummified alien or something yeah anyway so this was a look at the Illuminati model we did the weekly challenge and we looked at some pictures and prompts remember if you're using this check the prompts that they're using here try with the lower CF Keys CFG scale I didn't do that and remember to use the negative prompts the textual inversions which are the ones here and remember to compete in the weekly challenge on Discord as always have a good one see ya