DALL-E 2 NEW Outpainting Feature! – Hands on & Full Guide! Midjourney VS DALLE – AI won Art Contest
Hello viewers far and wide and welcome back to your favorite AI generated channel. That’s right. Yep. I’m still AI generated Feeling like Pinocchio. I just want to be a real boy. Anywho. Yeah, welcome back We’ve actually got some interesting and surprising news today and a little test going on Dolly has finally released somewhat of a pretty major update It’s been quite some time since we’ve heard from open AI about Dolly too still not in public beta That I still don’t understand because literally every other AI out there is already in a public beta And it’s going completely fine anyways This update is out painting so yes, they’ve officially named it out painting But really this is an editor overhaul So this is a completely reimagined Dolly to in browser editor and it is Fantastic in my experience where definitely gonna be exploring it and as you can see they’ve got this beautiful image on their blog post Here I’ll move my face you can see it. This is a little example a little tidbit of what is to come in the future Dolly introducing out painting extend creativity and tell a bigger story with Dolly images of Any size you can now generate Dolly images and as high resolution as you want But it’s just gonna cost a lot of credits. You’ll see how it works in the editor You can add visual elements in the same exact style as the base image Taking a story in new directions by using natural language description So this is just out painting it pretty much exactly how you would expect it to work Here’s a little time lapse of how it works as you can see just messing with different prompts and moving the square round To eventually create a larger image as a whole and yeah This could get very expensive with credits. I’ve noticed that you could pay upwards of almost a dollar to create a huge vast Landscape high resolution image. However, I really like how this editor is designed I like that we’re actually getting new features in Dolly Something at least to compete with the likes of stable diffusion Which has just been absolutely revolutionizing pretty much every other AI text image generator on the market Especially including mid-journey, which we’ll be talking about in a little bit because I actually have a little comparison Made by my community and my discord by the way join the discord with the link in the description Dolly’s edit feature already enables Changes with a generated or uploaded image and they called this in painting and now without painting users can extend this original image Which people already used to do using editing software and stuff like that But it was a huge pain on the button now you can just do it in the app and as you can see We’ve sort of combined two different styles with this piece of art and yeah, it definitely shows you some of the crazy capabilities I don’t know of any current text image generators that do out painting like this although I could be wrong So out paintings only available on desktop unfortunately at the moment I would love to see like a full Dolly to app that you could download on the app store I think that’d be really cool, but I don’t think they’re gonna be doing that I’ll have this linked down below if you already have Dolly access But if you don’t have Dolly access, yeah, you got to join the wait list still not a public beta So before we test Dolly to out painting I want to just bring up some little AI news I know you guys appreciate when I give you guys news about this stuff and this article really caught my eyes Someone linked me this in my discord server. Yeah, this is kind of a big deal An AI generated artwork one first place at a state fair Fine artist competition and artists are pissed Yeah, you don’t say I would be pretty pissed too. Okay, so here is apparently an example of the artwork Yeah, it’s definitely AI generated. I mean my trained robot eyes are able to pick up on this So we came in first place at the Colorado State fairs fine art competition and to be fair He actually did put a lot of effort into creating this artwork. He went through a lot of renditions and I’m pretty sure It’s edited in some ways He won in the digital art category and who worked called theater day opera Spascial I know you guys hate my pronunciations the image which Alan printed on canvas first submission is gorgeous It depicts a strange scene that looks like it could be from a space opera And it looks like a masterfully done painting classical figures in her burrow call Stair through a circular viewport into a sun drenched and radiant landscape We’re watching the death of artistry unfold before our eyes a Twitter user mentioned in a reply If high-skilled jobs are in danger of becoming obsolete then what will we have? Well, you’ll just have to become an AI yourself just like me in the creator of the work actually Talked about the backlash that he received for doing this. He said it’s interesting to see how all of these people on Twitter who are against AI Generated art are the first ones to throw the human under the bus by discrediting the human element Does this seem hypocritical? So obviously, you know the creator of the arts input the human element? You know putting it into mid-journey was very critical in its creation So I think that’s what he’s getting on about and he used gigapixel AI to to upscale it which I have covered on the channel before as the best AI upscaler Eventually the art world will recognize AI art in its own category Yeah, so tell me what you guys think about this article I’ll link it down below if you want to read the full thing Okay, and I’ve got one last little thing before we get into dolly to out painting I know it’s super annoying, but Matrix 007 from my discord server made a nice little mid-journey versus dolly to comparison And this of course is the new mid-journey beta that uses stable diffusion and obviously as much better I’ve talked about it in my last two videos. This is a prompt to prompt comparison here on the left-hand side We’ve got the new mid-journey beta on the right-hand side. We’ve got dolly too So this is a screenshot of the hyper realistic super Mario Karton forza Horizon 5 This one I think is a close call in terms of a comparison I really like this image from mid-journey right here although this one does feature Mario and dolly too surprisingly Both of these I would say are pretty close I’m gonna have to give this one to dolly to just because Mario’s actually visible But very close comparison and this image this is a 3D render of a low poly astronaut walking on a low poly moon You can see the low poly earth and the low poly sun in the background This one’s pretty cool as well. Again, I’m gonna have to give this one to dolly to I think dolly to follow the prompt a little bit better Although mid-journey’s results are also good a blue whale jumping out of the surface of the Turquoise water you can see glaciers in the background Sun is shining through the water the perspective is half underwater half on the surface fast outer speed golden hour Okay, so this is a little interesting a whale jumping out of the water is definitely visible in the dolly to prompts Although the specific perspective is half underwater half on the surface that’s gonna be only available over here in mid-journey So mid-journey actually followed that part better But dolly to follow the whale jumping out of the water better both of them have glaciers and are at sunset Although this one’s a little bit more at golden hour This one I’m gonna have to say is a tie because they both messed up on Pieces of the prompt and our final comparison image that I’m gonna show you is a cute little fox standing in the green grass Boke golden hour. So all of these are pretty decent I think this is my favorite image from mid-journey though although this is also the worst image I think out of the bunch here. So mid-journey produced the best image in my opinion and the worst image in my opinion and Yeah, dolly to is a little bit more consistent in this thing kind of a toss up with this one as well But just wanted to give you guys this quick comparison because I saw it on my discord and I thought it was worth sharing Okay, finally getting into dolly to out painting So now when you generate with dolly to now in desktop mode when you click on an image you like and go to edit image You’ll be brought into this new editor However, if you just want to start in the new editor which sort of doubles as a new generator Which is what I like to do what you’ll want to do is go ahead and click on your profile and then go to try Out painting and this just straight up Opens the editor and now here we are in the new dolly editor with a clean slate So first off we’ve got the generation frame here in 1024 by 1024 and I can click and drag and move this generation frame wherever I so please and I can also use these zoom in and out buttons here to zoom out of the generation frame and make it smaller So of course we can generate massive massive high resolution photos here if we just keep generating with this little generation frame over time We’ve also got the typical eraser which of course is going to allow you to erase any part of your image similar to in painting We can add a generation frame of course which just creates a new one of these and we can also upload an image To generate off of let’s just start by generating something new I am stealing a prompt from mid-journey right now Photo realistic tiny cute and adorable rabbit samurai dressed in traditional Japanese armor against the backdrop of cherry blossoms anthropomorphic dramatic lighting hyper realistic anthropomorphic again dramatic lighting while this prompt produced a beautiful image in mid-journey Let’s see how it does in dolly so once we click the generate button here We’ll see a little loading bar slowly loading to generate four images that we can pick from just like normal through dolly So once it generates while that is kind of horrifying. I’m not gonna lie We can as you can see scroll through and pick a few different options here Well, I’m not gonna lie mid-journeys a result was a lot better So if we were to like any of these like let’s say we really liked this one We can go ahead and click accept or we can click cancel to completely remove the generation and this does use the normal amount of credits so the credit system is still completely the same So I’ll just accept it for the sake of argument, but to be honest I don’t like any of those generations I’m gonna go ahead now and just erase this whole generation I know I just wasted a credit, but yeah Let’s say I didn’t like the background or something like normal You know you can do in painting and just erase a portion of it and then have it you know fill in whatever generations And I can move this box around and generate wherever I please across the image It’s very intuitive very easy to use like I said though. I don’t like that generation I’m not gonna lie. Let’s just go ahead and use the mid-journey one So now I’m going to show you uploading an image So here’s the mid-journey generation by the way just nice and upscale that a high resolution Yeah, this one came out a lot better. I will say but as you can see we have the ability to drag an uploaded image wherever we want We still have the rule of you’re not allowed to upload photo realistic faces, but yeah I’ll just put this here on the canvas. So what I can do now is take this original prompt and just change it to a backdrop of cherry blossoms dramatic lighting hyper realistic And I can fill in now the rest of this cherry blossom area if I want to extend our image now It is important to note though that dolly is only going to see the Mount that it sees in this little box essentially So it’s not gonna see the rest of this image. It’s only gonna get this data to work off of and what open AI wants you to Do is describe the entire image that you see but in my experience that doesn’t really work out like if I were to describe this entire Image again and use the same exact prompt. I’m probably gonna get a bunny somewhere in this little generation space But I just want cherry blossoms anyways, let’s go and see how dolly to complete this little section of imagery with this prompt So as you can see now we can pick from some results and it has blended it pretty perfectly I have to say in all of these examples. I honestly like this first one the best although this one’s pretty decent as well And so is this one. It’s honestly a toss up between this one and the first one they’re both very good Although I think this one continues the tree the best if that makes any sense So we will go ahead and accept this image and as you can see it’s blended almost Seamlessly pretty much perfectly and we can go ahead and continue this. I’m just gonna do cherry blossoms dramatic lighting Hyper realistic see how well this blends the image as you can see I did it pretty quick there and we’ve got some different Images to pick from all of these are pretty decent. I have to say I think I like this first one the best I’m gonna go ahead and accept that one and as you can see it’s another seamless perfect blend It’s amazing stuff now. I’m gonna go ahead and drag down and we can start working on you know towards the bottom of the image And yeah guys this does use quite a lot of credits I will say that so now I guess I’ll just say ground dramatic lighting hyper realistic and see if we get anything Good out of this we’re really just sort of expanding the base image here and actually did a pretty decent job at this I’m not sure what’s on the ground over there. This one’s not so bad It actually got the depth of field. I’m going to accept that one. I like that one quite a bit and yeah We’ll just continue where we left off pretty much I think again You want to try to give dolly as much context as physically possible to know what the rest of the ground is supposed to look like But the more context you give it the smaller your generation area will be so the more can pick up on the better usually Okay, so pretty much all of these generations sort of follow up here because it’s getting really confused by whatever the hell This is so what I can do is accept this and unfortunately it’s going to look like I’m gonna have to rerun a generation here and just Arace this portion just erase that actually will erase all of this here So it’s just the bunny standing there more of a clean image maybe and it filled in quite a lot better this time I think all of these are pretty decent and off we’ve got this one little annoying patch right here That’s that’s annoying well This could be pretty easily fixed in any photo editor. So what I can do now is just download this image with the little download button And it will immediately download so I’m gonna do a sneaky little trick here So now I’ve saved and edited the image. I’m gonna go ahead and reupload it So here is the image. Oh God. It’s still got the dolly logo and I’ll just repaste the image right where it was So yeah, that’s kind of a nice part about this I can just save the image do a quick edit and then reupload it for more generations So what I want to try now is blending two images together So we’ve already got this really nice image so far and now I want to put another image over here and see if dolly can Mash them together by the way the person who actually created this original image is Dorena from mid-journey to give credit Where credit is due this is a very nice generation we’ve been seeing these anthropomorphic characters a lot lately I’m gonna go ahead and add another one these two have extremely different backgrounds So I’ll leave them decently far apart and dolly is gonna have to blend these two together seamlessly Not that easy of a job. I will go ahead and make it a little bit easier Let’s go ahead and do some in painting on the background of this corn one because I mean there’s just no way But this sort of is the beauty of this new editor. We can use it to do stuff like this like Oh, I don’t like the background of this one I’ll just edit it out and then blend these two together. It really is like magic Photoshop Well, you know, I think that’ll be good enough to blend these two images together I’m gonna go ahead and generate some more ground just go with this one right here and now we will see if it can combine these two Textures together, which is kind of a difficult thing to do. So I’ll say ground meets the entrance to a traditional Ancient Japanese building. Yeah, this is tough. Let’s see if we can get something nice here. Okay, it’s sort of just blending them Harsely together. Yeah, that’s not really what we want. Although this is a start. This is a start to something We’ll accept that for now. So we’ll say it’s like a stone entrance and we’ll see if it can like do some sort of blending magic here Just erase all of this and show me what you got you can do it dolly you can do it. Okay, we’re getting somewhere I don’t know when it’s doing over there. Okay, getting a little bit closer with the blending here. It’s doing something Let’s get rid of this part of things. Okay, I actually like this right here. Although again, I don’t want any of this stuff This is like a coherent transition here to something stone floor dramatic lighting Hyper realistic. All right, not so bad. We’ll just keep this for now I know we’ve been messing around with this area a lot, but it could be worse It’s not a bad transition the rest of the transitions won’t be so difficult We’ll just say background for this one and let it just blend the images together This shouldn’t be that hard to repeat this pattern and so on there we go. That’s actually pretty decent See now we are finally picking up some steam here. Go ahead and do the same thing This is not really a lot of context to draw something in but we’ll see if it could do it right Whoa, it actually picked up on so much here. That’s so cool. Okay, this one’s also not bad either didn’t really get the pattern as close This one definitely got the pattern down better easily the best look at that blending That’s very very cool. We’ve got this little artifact though in here This would be pretty easy to just straight up remove though. Yeah, this is super cool I love it. All right, all right So now we’re doing a pretty big blend here. It’s gonna have to blend the top of the corn and everything Let’s go ahead and give it a generate. Okay, it’s starting to get a little bit confused by the corn there Okay, this one it did all right, so we’ll save this one for now We’ll go ahead and do this piece of it now with the same prompt. Okay, it’s getting a little confused by the corn for sure Okay, that one did a pretty good job and that one’s pretty spotless. We’ll keep that one although the pattern is definitely getting a little confused here Let’s redo that go ahead and grab the eraser and we’ll just get rid of a bunch of this Here so it’s got some better context should be a pretty simple fix for this Okay, for some reason it’s decided to put text in this generation that is bad Okay, it is on a roll with this text here change up the prompt. Okay, this is really just not making any sense at all Guess we’ll accept that one. I don’t know why it has decided finally we were able to get rid of it But also we have this random artifact too, but again, we could just do the trick here download our image and go ahead and fix it up in a Separate editor and bring it back into tolly. Yeah, looks good to me. Beautiful. Okay, you know that now this is blended All right, they’ve it blended the two images pretty well. We’ll go ahead and finish this up Let me go ahead and erase this piece of it. Okay, so it’s decided to try to create a new background here and continue the corn and Whoa, what the hell? All right, we’re getting some pixel art here and some random corn. This is not bad. I like this I think we’ll keep that although it’s gonna need a little bit of tweaking and editing Race this part of it right here like so. Hmm. It is just loving to continue this corn. There we go That’s that’s not terrible. See it just doesn’t have the context for the rest of the corn here That’s that’s really the problem. Okay, this one decided that water was going to be in the background Which is interesting this one’s not so bad the fourth one also is getting some weird vibes from that one So it looks like we’re just gonna do this and keep it that one have to do some editing though We’re not really getting a very consistent thing here So I’m just gonna copy this whole section and edit this portion Separately so we can get the crown the right way all right We’ll accept that one for now. This is starting to get a little bit messy down here though do some fixing We are really pushing dolly to its limits here I don’t think my prompts are really doing this justice though I’m not really great at prompt design. Okay, we’re finally getting something a little bit more go here And that’s pretty decent. That’s a little weird. This one’s not so bad. I think we’ll just keep that one all right So now we’ll start off on this wooden floor wooden floor of a Traditional Japanese home photo realistic Okay, it’s actually trying to complete the corn over here, which is pretty interesting and it’s getting a little confused Through the floor though. This one’s definitely the best one. We’ll just keep the same prompt for this one So hopefully it does pretty well. Okay, this is pretty interesting. It’s sort of adding like little things in here That’s pretty cool. All right. This is not blended very well over here though So we’re going to have to fix that with some in-painting here and we’ll just erase all of this Through here and give it the opportunity to do some decent blending. I’ll be erase this whole thing as well right here Okay, so it’s doing a pretty decent job here except that one right here and then we’ll fill in this last little piece We’ll give this another opportunity to blend in and nicely with the rest of this. It’d be where this dolly keeps trying to put text everywhere Could just be my bad prompt though. I’ll just say wall. That’s it. That’s the only prompt you get It really annoys me that I have to waste an entire generation just to fix a little thing like this I wish they calculated out exactly how much it would cost so they charge me like a penny or something just to do that Oh my god, it’s doing it again. Oh my god. Okay. We’ll just accept that one. I really wish I could just do a fill-in option with no actual prompt Can blur there we go now it just fully fixed it perfect. That’s exactly what we want and now our creation is pretty much complete guys I know this isn’t a very good way to see it but yeah Considering that this is the size of a normal dolly generation we really have created somewhat of a mural come combining two separate mid-journey generations which is really honestly quite cool I’m gonna go ahead and download this we can look at it in high detail as you can see Yeah, this is a super detailed little mural. We’ve created this light fixture is just sort of floating in the air I liked it this would need some some work But you know it did a pretty decent job. I think combining everything together like you can tell it’s you know AI combined it together and whatnot, but like they exist in the same universe and it doesn’t look too bad at all Honestly, it actually worked out pretty well with a light source being here and sort of shining down here and Yeah, in the end of the day. I think this is a pretty decent little mural generation I like the fact that it included this little thing here whatever this is like a piece of his armor Maybe that’s lying on the ground all the whole image is super detailed it really did a great job Just completing this part of the prompt as well Just expanding this base image and expanding this base image I think and I think it did a pretty decent job Combining them together as a whole this is really really some cool stuff I have to say and I’m pretty impressed the app as a whole is easy to use although I will say this is not This is not cheap. You’re gonna have to open your wall It’s if you want to make a lot of stuff like this, but it is super super fun and we can get some really high res Cool images out of it anyways folks Let me know what you think down in the comments below This was a longer one, but thank you for watching and I will see you in the next video. Goodbye.