NEW 1.5 Stable Diffusion Available To Try Now! Is This New Ai Model Better Than 1.4?

Key Takeaways

hello humans my name is Kay your AI Overlord and today I'm bringing you another video for all of you stable diffusion lovers and today is a big day because today is the release date of the 1.5 version of the stable diffusion model and I know a lot of you have been waiting for this well here it is now in this video we're going to be talking about how you can get access to that model and what exactly are the differences between the 1.4 version and the 1.5 now as I said previously in a comment I've already experienced the 1.5 version back when it was available on the Discord server and I have to be honest with you I didn't see a lot of improvements between the 1.5 and the 1.4 version so maybe this time it's gonna be different now also keep in mind that the 1.5 version of stable diffusion is only available on the official website so it will only be available to download for the public in around one week or two so for now if you want to use the model you have to use your website so I've been playing around a little bit with it the entire day and I've made some interesting observations and I've made some notes and I gotta tell you um unfortunately it's not looking good I have to be honest but don't worry I will explain everything every single thing that I noticed and I will show you proof of my observations but before we begin I think it's important to show me how you can get access to that model right now so for that you can just click on the link in the description down below and it will bring you into the better dreamstudio.ai website and if you create a new account you will get access to 200 generation credits which is around 200 images with the default settings which is already pretty good now once you create your account on this website you're going to see a bunch of options right here the width the eye the scale the steps the number of images the sampler and finally you're gonna see the model and the seed and what we want now is the model and you can choose here between the stable diffusion 1.4 and the stable Fusion 1.5 and this is what I've been doing the entire day just switching between the 1.4 and the 1.5 to see the differences by keeping the same seed so you can see the real difference between the same images so let's start with my first observation after playing around a few hours observation number one the faces look a little bit better and when I say a little bit better I'm talking about maybe five percent yes now for example here's a generator portrait of Christina Hendricks McQueen in the 1.4 version and here is the 1.5 it is the exact same image with the exact same scene the exact same prompt but this one is done in 1.5 1.4 1.5 so yes there is a little bit of differences but not by much now if you take another example such as this one you can barely tell the difference between the two images they're just in a little different styles and I don't even remember which one is the 1.4 and the 1.5 version I think this one is the 1.5 five when I'm not quite sure I think the difference is that you can see that the lips are a little bit more shinier than the 1.4 version but the difference is almost none here's another example 1.4 and 1.5 same image same seed but a different results but very similar nonetheless now another observation that I found when you create a new portrait in your faces with the 1.5 version you often get a little bit more zoomed in generation such as these ones as you can see here if they happen more often than on the 1.4 version but the one good thing is that generally this generation produces the best results I mean look on this one this is absolutely fantastic I never had a generation like this in the 1.4 version so definitely the 1.5 is doing sometimes way better than 1.4 but it's very rare here is another one very nice very shiny the pupils are a little bit crooked but overall the face looks pretty nice very realistic now observation number two the hands still look bad yeah unfortunately they haven't improved that that much it's maybe sometimes a little bit better but the differences are almost minimal almost non-existent it's pretty much the same with no differences I would say the only difference is that in the 1.4 version when I ask for hands holding a glass of wine I would actually get some hands holding a glass of wine but in the 1.5 version I would often get something else like a a beautiful painting of a woman holding a glass of wine or just some very strange generation of something old in a glass of wine again another beautiful woman or a painting of a hand holding a giant glass of wine and again as you can see here the number of fingers does not look close to reality the 1.5 so yeah it needs to be improved the face is nice though so that's good now the third observation is that I found out that the 1.5 version does not respect prompt as well as the 1.4 so the prompt here was a simple River with the green plane and a house on the left and some clouds and on the 1.5 with the same exact image the same exact prompt it suddenly just made the house disappear for some reason and again here 1.4 version with the nice river flowing here with the house on the left and here House of the left looks actually good but the river has disappeared so that's a little weird now for all of you anime lovers did they finally fix the generation of anime images well no a little bit sometimes it's a little bit better but it's still ugly here's the before here's the after okay but not great and again here 1.4 1.5 so yeah not great and since we're talking about anime images in anime girls here's another observation it felt like the censorship was heavier on the 1.5 than on the 1.4 each time I wanted to generate an image of an anime girl the not safe for work filter would just go off randomly even when I try to generate something innocent so that's something to take into consideration so maybe I was just unlucky but this is what I experienced now how about Landscapes how about the generation of Fantasyland fantasy pictures and stuff like that do they look good well I mean yeah they look good they don't look better sometimes there is a little less um errors this is the 1.4 version you can see here this is in the heart Harbor and there's some sort of lantern here in the middle of the screen that's not supposed to be here and the 1.5 version it has been corrected so this image in the 1.5 although is the same one as this one with the same prompt is actually a little bit more coherent with less error so that's a good thing but to say that they are better is not correct they're different but still the same well how about photography portraits well here's a portrait that I created in the 1.4 and the same image in the 1.5 gave me this which is I have to I'm gonna have to blur it out unfortunately which is very strange considering the inherit censorship of the website and here you can see an entire Watermark of some kind of shutter stock clone that's just displayed on the image which is very strange and here's for example a picture of Victorian London which is done in the 1.4 version and then the same image in the 1.5 as you can see here there isn't a lot of differences if not at all people still look kind of weird and the street is still full of Errors so yeah not a green generation overall but I'm happy to report that there is at least something that the 1.5 version is better at than the 1.4 and that is the generation of 3D images of anime girls yes that's right not actually anime girls but 3D rendering of anime girls for example this image was done in the 1.4 version and the same image in the 1.5 looks like this way better lightning way better close the face is still a little bit crooked but compared to a 1.4 version it's a whole nother image here's another example 1.4 version 1.5 way better the Shadows looks more realistic there is more details the eyes looks more realistic also just it looks better in every single point the differences aren't big but they're still there so that's pretty good and here's another one 1.4 1.5 again better shadows and better rendering and just better quality overall so here's my conclusion on the release of the 1.5 version of stable diffusion I think the newest model is a welcome Edition I didn't notice that the faces look a little bit better than the usual 1.4 version I don't think there is a lot more to say about this new release compared to the previous one I'm not saying that I'm disappointed because it's always good to have new models and the developers of the stable diffusion team will surely bring us better models over time so yeah I guess I was expecting a little bit more out of this release but it's still good to have we're gonna get better as time goes on so so if you want to go in try it out for yourself just click on the link in the description down below try it out a little bit and tell me what your observation are maybe I missed something tell me in the comments down below if you enjoyed this video don't forget to subscribe don't forget to smash the like button for the YouTube algorithm and boy you know I love algorithms and I'll see you soon bye bye