NEW Photorealism Model

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
hello my friends today we're going to look at a stable Fusion model that will give much much better photorealism while sdxl has been great for a lot of images but it has a lag is well photorealism especially in people humans in general so let's see if this model can sort that and in my opinion I think it's pretty good oh and by the way to the person who cut in front of me in line I'm after you now remember guys I do the research so you don't have to so if you want to help me out feel free to like subscribe put a comment down below it's really going to help me out with the algorithm but it's okay if you just don't watch too hey so it's probably about time to start changing this background now we all have a lot of images from stable Fusion 1.5 and 1.4 but anyway let's look at some sdxl images and isn't this just fantastic so this is an image of some sort of a Viking we have a post-apocalyptic man with a skull mask and a sort of a dystopian Society here we have a woman in the jungle wearing some beautiful blue flowers there here's some kind of cyberpunk vibe some scientist with mechanical eyes I just love how the hair flows in this image and catches the light here's the black and white photo of an elderly woman looks very nice here's uh Superman looks a little bit like Henry Cavill possibly cute little frog we have a beautiful photo a snow leopard and I mean if I didn't know this was an AI generated image I would probably think this was real something I've noticed is that animals stuff like that they look very good fantastic even the fur is great the details here on the ground are fantastic you have some branches here that actually look very realistic so this is in my opinion very good now skin in general leaves a little bit to be desired I mean this is pretty good I wouldn't say it's 100 and you can especially see it on women for example where a lot of the time is very smoothed out or very like makeup shiny kind of effect they're still pretty good and in this photo here will photo the generated image here this could be taken straight out of a cinematic scene from like a TV show or a movie or something I find this very nice here's an example of of a car and all this comes from a custom stable Fusion sdxl model and that one is Juggernaut XL to be specific and Juggernaut was a model made for 1.5 that a lot of people have been using and is very happy with and I downloaded this a couple of days ago well not a couple of days actually yesterday because it was uploaded yesterday and I have been playing around with it and I found it's uh it's actually pretty good we have some examples here and it says it's supposed to be a photorealistic model however it's an early release they're expecting to make it better but still you know it's it's very good and the custom models in general really start to outshine the base models same as they did in stable Fusion 1.4 1.5 2.0 2.1 Etc and I expect this to keep happening with sdxl so all you need to do is download this model here there's a vae that you can download here it's not required but is recommended so those two files and there's also a cinematic xlora you can get from here and once you have those three files depending on what UI you're using right now the folders are pretty similar but say you're in focus for example you have checkpoints does it this is where the save tensor the main file goes the general Excel then you have the vae you put your vae there then you have your loras you can put your loras there this is similar for automatic 11 11 or St neck comfy Etc they all have a similar structure of how you're supposed to of where you're supposed to put your model and once you've downloaded that and pull them right in restart your stable Fusion go into whatever it is you're using let's have a raw candid cinematic scene of a Viking Warrior and let's generate two images here we are inside Focus which is a very easy to use interface without a lot of features now you can check my previous video on that app and the top right corner now if you just want to get started with stable fusion and if you're new to all of this I can highly recommend it for beginners and we can see here coming in live now are Viking Warrior so we got two of them here's the first one and just look at how fantastic this image was with a simple prompt raw candid cinematic scene of a Viking Warrior a negative prompt but just smooth plastic nothing fancy for a lot of the images you don't even need to use negative prompts nowadays and the second one that we got was this one which is also fairly okay but I think that this one was uh just a tiny bit better and this was the first generation we got while recording so this was non-cherry-picked and I can highly recommend this model as of right now now there aren't a lot of sdxl custom models yet but we will there are a few and you can find the link in the description below so go check it out and see what other models you can find let's generate something else here so let's keep a raw candid cinematic scene here and let's um take off a Sci-Fi space ship in the future let's see what that gets us we've got two images here they're fairly dark where I find that this one was quite cool looks pretty good and actually looks uh looks very cinematic now remember that if you haven't used a new model before you need to activate it depending on your UI it looks different for Focus you're going to go into advanced just click this little checkbox down here and then select your base model up here and if you want to use the Laura you can select that in here but for any user interface whether it be comfy or automatic or whatever it's fairly easy to change what model you're using Raw all counting cinematic scene of a cat and a street reigning neon signs cyberpunk Blade Runner and we see we can see the outlines of our little cat can get in there and uh boom let's see if we get some Happy accidents with this image that's what I love about generative AI all the little happy accidents and the Beautiful Generations that just appear now when you're working with control net and stuff like that you always kind of know like the direction where you're going but in a user interface like this where everything is more or less random it's great fun oh man these are fabulous I love these amazing images like I said previously one of those animals it's very photorealistic as that's not specific for Juggernaut but for sdxl in in general but um man for sure these are great hey I hope you enjoyed this let me know your best model tips in the comments below and as always have a good one see ya
Info
Channel: Sebastian Kamph
Views: 18,165
Rating: undefined out of 5
Keywords:
Id: uKmcYDpQqrA
Channel Id: undefined
Length: 8min 8sec (488 seconds)
Published: Fri Aug 25 2023
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.