OpenAI just dropped a new image generation model with a completely redesigned interface—and after hours of head-to-head testing against Google's Nano Banana Pro, the results surprised me.
So, OpenAI just released their brand new image generation model and a sleek, beautiful looking new interface inside of chat GBT for image generation. Now, I've been testing this out for a few hours now and I've come up with some pretty surprising results that I want to share with all of you and we're going to put the new Chat GPT image generation tool head-to-head against Google's Nano Banana, which is considered one of the best generation tools out there for images. and we're going to see if ChatGpt has overtaken it in terms of image generation. So to access this new tool, just come over to chatgpt. com and click on the left here where it says images and that opens up this beautiful new tab where you can generate any images that you want. ChatGPD is calling this simply chat GPT images. So up here at the top you can of course type in any prompt that you want and it will generate that image for you. But they also have down here this section which is where you can try these different styles on an image. So, you can select any one of these, like I'll select this one right here. It will then ask you to choose a recent photo that you uploaded or choose a new photograph. So, I'm going to choose this one of me. And that's going to open up a new chat where it autopopulates the prompt attached to this image. And then, of course, it'll start working on creating that image. And you'll notice that this is substantially faster than past models of chat GBT of their image generation tools. This used to be pretty slow, but this new image generator and my testing is much faster. So, this is what it came up with. I think it's a pretty good image. It took just over a minute to generate this, which is faster than the older model. So, very impressive to see this. Now, coming back over here to images, you'll also notice that in addition to all of these styles, we also have this section called discover something new where you can also do some fun stuff in here as well. So you can turn yourself into the girl with a pearl earring, restore some old photographs, create a professional product photo, create a holiday card. So for instance, right here, I made myself into the girl with a pearl earring. A little bit creepy, but hey, it's a little bit fun. And I think that's what's really stands out to me about this section is that this is going to be a lot more accessible to the average consumer. The average consumer doesn't know about prompting. They don't want to have to search through Reddit or X for the perfect prompt to generate the photos that they want or to create these artistic fun styles. They just want to come in here and have some fun and this really enables them to be able to do that. So, for that alone, I think this is going to really give Chad GPT a major leg up and is going to have a lot of people coming over here excited to use it just because it is so easy to kind of jump in here and start having fun with image generation and editing. Now, like I said, OpenAI didn't just update the user interface. They also updated the underlying image generation model. It used to be GPT image 1. Now it is GPT image 1. 5. Now, they didn't promise that it's going to make more highquality images, but they said it's going to be faster, which we've already seen, more detail preservation, precise editing, and stronger instruction following. So, let's go ahead and test that by pitting this head-to-head against Nano Banana Pro. Okay, so for this first test, I want to do something pretty simple. I'm not going to upload any reference photo. have it edit a photo. It's just a simple prompt asking it to create an image of a cat. And I'll go ahead and start the chat GPT one first and then the nano banana. And I really want to see which one is faster at image generation here. All right, so nano banana is already done. Chat GPT is on the way, but it's not finished yet. So, we can see right here that this is the image of the cat. I asked for a cat sunbathing in Tuscanyany. Uh, this looks like an amazing photograph honestly or amazing image it was able to generate. And then finally, the one from chat GPD is finished as well. So, a few seconds later, not terrible. A lot faster than it used to be, but not at the same speed as Nano Banana. And just looking at these two images, I got to say I like the one from Nano Banana better. It looks a little bit more realistic to me. This one is definitely cute. It's definitely a great image, but I think this one has a slight win to it. As I go through these, let me know in the comments section down below, guys, if you find that one image generation model looks better than the other one in your opinion. Um, I'm going to share mine as I go along as well. Now, one of the first things I wanted to test with the new Chat GBT images was how good it was at editing images. So over here, I took the same image of me, gave it to both models, told both them to change my shirt to blue, and these were the results. So this is the one from Gemini, and Chat GBT. And I got to say, both of them did a great job. However, I would give a slight win to Gemini. The shirt just has a little bit more texture. The one from Chat GBT just feels a little bit flat, and it even just makes my face look a little bit sort of like flat as well. So, it definitely made some changes to my face, it looks like. Or it could just be the coloring, but I don't think so. I think
Gemini gave a slight win to this, although Chat TBD did a fine job as well. But then I wanted to try something a little bit more complicated. So, I gave both of them this image of me driving a boat and told them to change up the photo so it looks like I'm in the cockpit of a spaceship flying it. And these are the results. And honestly, I'm not 100% sure which one of these I like better. Looking at the two of them, I like my skin tones in the chatbt one better. This one left me a little bit red. I guess that's the way I look in the original photograph, right? A little bit red, but this changed the coloring to sort of match the image a little bit better. So, it works like it's a little bit of a cooler color. And I think it just made a really better looking cockpit. This one still left a lot of the white from the original photo and like the boat. And in fact, now they look at it, you can see the railing from the boat as well. This one looks like I'm actually inside a cockpit a lot more than the one that was created by Gemini. In addition to that, the spaceships in the background look a little bit cooler in my opinion. These ones look a little bit more cartoonish. So yeah, for this one, I would say that Chad GBT actually won, but it is pretty close. So let me know in the comments section what you guys think. Now, for this next head-to-head, I wanted to test two different things. one, consistency of characters, and two, its ability to put text on images, which is something that Nano Banana has been crushing. And so, the way I usually use this is I have Gemini or Nano Banana Pro help me with my YouTube thumbnails. So, over here, I gave both them the exact same prompt and the same image, and these are the results. So, first of all, with Chat GPT, I'll say it did a great job of putting text on this image. This is something it's struggled with in the past and here it did a fantastic job. So did Nano Banana Pro, but there's no surprise there. However, for character consistency, I would have to say that ChachiD did a pretty bad job. I think maybe some of you at home will think this sort of looks like me, but I think this person's neck is way too wide. It doesn't really look like me. This one on the left that Gemini created looks like me. That is me. That's a really good image of me. So, of these two, I would say that in terms of consistency of character, Gemini is doing a much better job. Now, I wanted to be fair, so I went ahead and tried the prompt again, just changed it slightly, and I think it did a much better job with this rendition than it did with the last one. However, I will say that Gemini still looks more like me than Chat GBD does. But I wanted to push the boundaries of text to images, not just a few words like we had in the last example. So, I told both of them to create an infographic for me that covers the best AI tools. And here are the two different results that they each came up with. So, the one from Chat GBT is fine. It's a little bit simple. There's just six of them. And I actually wouldn't even call this that simple until you compare it to the one from Gemini, which has a ton more detail. It breaks it down into five different sections, goes over the best tool in each section. There are a lot more words here. So yeah, the one from Chat GBT is fine, but the one from Gemini is just, you know, it just has that advantage, just better than the one from Chat GBT. Now, something I use Gemini for all the time is I upload a PDF and ask it to create an infographic based on that PDF. Gemini always does an excellent job at that. Chat GBT failed at that colossally. This information inside this graphic has absolutely nothing to do with what was inside the PDF. take that with a grain of salt because I don't know if ChachiBT ever even said that this was a possibility, but I figured I would just throw it in there. Now, for this next test, I wanted to see which model was better at really understanding, I guess you could call it scale. So, this is an image I had created in Nano Banana in the past. I wanted like a giant image of me walking down the street of New York City, and it did a pretty good job. My foot here looks a little bit weird, but all the like heights and the ratios seem to be correct. Right, we have smaller people here now. Well, actually the people in the cars are normal size. I'm just gigantic. Now over here, the same prompt in chat GBT failed. You see, some people are large and some people are tiny. So, the scale is just way off here. So, I reprompted it and it still couldn't get it right. We have some normal people over here and again some tiny people right here. And then I'm standing on the car sort of awkwardly. So, neither of these did a great job, but in terms of scale, Gemini definitely did a much better job. Now, not that long ago, I created this image inside of Nano Banana of myself with the Joker on the set of Batman. I think this photo looks absolutely awesome. My hair looks a little bit weird, but overall, this photo's abs image is absolutely awesome. I asked Chat GBT to do the same thing, and it was rendering an amazing looking image, and then it just stopped and said that it wasn't allowed to. it didn't follow their content policy. So, that's
a shame. But, in the split second I saw it, it was an amazing looking image for whatever that's worth. For this next one, I had them generate images of the Brooklyn Bridge on the 4th of July with fireworks over the Statue of Liberty. And I have thoughts on both of these. So, this is the one from Gemini. And if you look at all the people's faces, a lot of them are kind of blurred out. There's not a ton of detail in their faces. However, there are a lot of people here. What I do appreciate this about this is that it got sort of the realworld scale and aspects of this quite well. Like this is what the Brooklyn Bridge looks like. The Statue of Liberty is over there. And if you compare that to the one from Chat GBT, it just kind of got it really wrong here. The Statue of Liberty is not this close to the Brooklyn Bridge. And here the people are like sort of on this second bridge that isn't connected to the first bridge. It's a little bit awkward and weird. However, the people here, even though you only see the backs of their heads, seem to be a lot more detailed, although almost cartoonish. I would say, in my opinion, Geminis's is better, even though the people are a little bit blurred, but it sort of works for this because this is a nighttime shot. They're not the main focus of the image. Like that's the bridge and the fireworks. So, this sort of works for me a lot better than this one, which looks a lot more fake to me. So, as you can see, so far I'm really feeling like Gemini is a better image generation model, but I wanted to really find is there something that chatbt is better at. So, remember we generated this image before, this really dramatic image of me that I thought was pretty good. And when we generated it, it actually showed us the prompt. So, I put that same prompt into Gemini. And I got to say, Gemini actually did a better job at ChatGpt's own prompt. This one looks a lot more like me. You can see a lot more detail. I think it look really fierce here where in this one it just kind of it's very dark. It's actually hard to see me. So even at Chat GB's own game, so to speak, I still think Gemini is winning. Now another place for me where Chat GBT failed honestly is in some of the image editing. So they have this really cool feature where you can open up any of your images and go ahead and select any area or just tell it exactly what edits you wanted to make. So I can go ahead and like select let's say I wanted to get rid of this book. I can highlight this and say like get rid of this book. So what I did was I actually asked it to remove item number one on this list because it's incorrect. Chat GBT or OpenAI never said that GPT image or image 1. 5 has enhanced image quality. So that was actually wrong. So I asked it to remove it and it just did not understand what I was talking about at all. So I download the image uploaded it to Gemini and it understood exactly what I was talking about. It not only removed item number one but then renumbered the other ones so that it wouldn't start at number two. So it did an excellent job where chat GBT just failed. Now all of this is still very initial testing. As time goes on we might learn better how to prompt chat GBT and get better results out of it. And I want to give them credit where credit's due. So, for instance, this is an image that I had generated inside of Gemini a while ago. I really like this image. I think it looks really cool. I sent the same exact prompt to the new image 1. 5. And these are the results. The robot it came up with in the middle looks freaking scary. Like that is a really cool looking robot. Very different from the one that Gemini generated. So, I think this is super awesome. But you'll notice it sort of has like this yellow hue to it. And I was looking on Twitter and a lot of other people were seeing this same yellow hue as well. It seems to default to that and it's sort of like a strange color scheme that's showing up a lot and there's a couple of small details in the background that I don't love, but yeah, it's still a really good image that it's able to make. In addition to that, I will say that I think the images it produces are better than the last model. So, this I produced with image one a while back and then today I gave it the same prompt and I think this image looks better. I think there's a lot more detail to it. I think it's just overall a better image. So, I think there are definite improvements. I think everything they're claiming with the stronger instructions, the precise editing, detail preservation, and four times faster is accurate. So, I think for a lot of users why this is going to be a gamecher is because a lot of people use chat GBT and they're very happy with it as a chat model. But I know a lot of people see what's going on at Gemini and they're sort of jealous that it's able to produce such good images. So I think for a lot of consumers, this is going to be perfect for them. They don't need something as powerful as Nano Banana Pro. They're going to be happy with this. They're going to have fun coming in here using these different styles on images and this section right here and being able to edit images. This is going to be perfect for them. And if this is what 1. 5 looks like, then when they finally release image 2, that could be a total game changer. that could be an incredible model. But for myself, I'm
going to continue to use Gemini for Nano Banana Pro anytime I have to generate images. It just feels like it does a better job of following directions. Another thing I didn't mention in this video is I've been struggling to get Chat GBT to actually render images in the correct aspect ratio. Like for instance, the thumbnails. I tell it it's the thumbnail. it should produce an image that is the correct dimensions and it's not able to do that uh without a lot of prompting where Gemini just does that for me automatically when I ask it to create a thumbnail for me. So, it's just a little bit better in those different regards. But what do you guys think? Let me know in the comments section down below. If you've tried out the new image 1. 5, what are your thoughts on it? if you've compared it to Gemini, if you like Gemini better, and if there's some things that you're going to use both of them for, what are you and how are you going to use them? Then otherwise, if you enjoy this video, make sure to subscribe. I'm always testing out these new AI updates that come out to Gemini or Chat GBT. I try to compare them, make videos for you guys so you can decide which tool you actually sign up and use yourself. So, if you appreciate that, make sure to subscribe for future updates, and then I'll see you in the next video. Bye for now.
