Bagel: NEW Chinese AI DESTROYS GPT4o? (FREE!)

We're going to be testing out Bagel, which is a new Chinese AI that's fully free, fully open source. It was released by Dance, which actually owns Tik Tok. As you can see right here, they say Bagel, the open- source unified multimodal model. So, it can also generate images. For example, here's a demo of it. So, let's play this and see what we get. And you can see examples of what it can do. So, it can actually read images. It can tell you what images are. It can generate images as well. And then I believe it can also generate gifts as

well. So we'll be testing out sidebyside versus chat GPT40. And it's a very similar model in terms of parameters for example. I've actually got a cheat sheet inside the AI success lab which is completely free. Runs you through exactly what is what you get etc. how you can use it. And it's also got all the links to what we're talking about today. But essentially if we look at bagel you can use this for image generation as you can see right here. So you say tell me about this picture and then it will tell you exactly what it is. Then you can ask about the artist etc. You can use this for image generation and also you can use this for editing images. So for example here's a picture that's been plugged into bagel and then the user has said he squatted down and touched the dog's head and then actually edited the image and has a picture of him squatting down and painting the dog's head. Same for this. So, for example, we could take a picture of the Mona Lisa, change to 3D animated style, and then you've got like a modern image of the Mona Lisa in 3D, kind of like a cartoon. Here's another one. This is quite interesting. So, you can actually use navigation, right? So, it says after 40 seconds, move forward. And then you've actually got a GIF that was created based on the image plugged in right here. So, it can do a lot of different things. Here's another example. And if you want to get started, basically you can go and preview it at demo. bagelai. org. It is open source as well, so you can run it locally. You can see an example of the GitHub page right here. Examples of what you can do, etc. You can add emotions to images, all sorts of stuff. Pretty cool. Let's have

a go now. So, let's take a picture of me and then we'll see if we can actually change the image and that sort of thing. So just going to download an image for me and then we're going to upload this into bagel. You can start a new conversation. You can actually change the language here as well. So you can change what language you have and also the image output and you can switch on think mode if you want to as well. And then you can switch between different parameters for the image as well. I'm just going to leave those as default. And then we're going to open up a example image and we'll say okay change this image. though it's a helicopter, not a plane in the background. See if we can change that now. So, it does take a while to

process. What we can do in the meantime is we can actually open up chat GPT and test exactly the same prompt inside chat GPT 40. So, we've got 40 loaded over here, which is kind of like the comparative model to Bagel. We'll plug in the image and then we're going to use exactly the same process right here. And there we go. So, we actually have the image in the background. Doesn't look great to be honest with you. Let's open this up. It's okay. It's not bad. Let's compare the original image here. We'll compare these side by side. So, this is the original image with the plane in the background, many stairs, etc. Little bit blurriier when you upload it. And then if we have a look here, it's very low res, but it has done the job. It has replaced the image with a helicopter as you can see, which is pretty cool. Now if we have a look at chat GPT let's see what chat GPT did side by side. So this is the image afterwards. The resolution is much better on chat GPT. If you actually have a look at this like the face is so weird the way it's changed the face looks very much like AI generated when I compare them side by side. Let's have a look here. So this is Chat GBT and this is Bagel. Honestly, I'm going to say chat wins just for quality. Let's open. Let's download this and see if it comes out better quality when you download it. Here we go. Yeah, it's still pretty weird, isn't it? And then if you look here, super fuzzy on the body, but at least it didn't manipulate the head. However, chat wins by a long way right there. Let's test another

thing out now. So, I'm going to say okay, new conversation. Upload the image and we'll say describe this image. Definitely takes a long time to get back to us inside chat GPT as well. I like the fact that you can just copy and paste an image directly in there. So, let's say describe this. If you actually go to a new conversation inside bagel, you can't copy and paste the image from what I can see. So, it's not chat cheapy in that way. But still it's free and it's open source. So I suppose that's what you get. Let's have a look at some other examples whilst waiting for this. I think it's drunk. So we've said describe this image with the image of me stepping out the plane blah blah. And then it's come back to us with this which is an image. Maybe we have to change the language output. Let's try that again. So if we have a look at chat GBT's output here. It said the image shows a well-dressed man confidently stepping out of a sleek black helicopter. Here's the details. Here's what it did, etc. And it was super fast in the response as well. Although to be fair, if you're using language output on bagel as well, does seem to come back to us with a much better image. So, for example, here the image depicts a man walking down a set of stairs from a small private jet. The man dress. Yeah. Okay. So, it did a pretty good job there. I would say on par with chat GPT. It's just chat cheap. Now we're going to go back to image output and I'm going to say change this to an image with the man dressed as an astronaut stepping out a UFO spaceship. And we'll switch to image output. We can put thinking mode on as well. But yeah, I got to agree. I think the lighting here is not that great on the images that we're getting back from by dance. We'll test this out. Here's the image of the astronaut. It actually looks a lot better to be fair. Look at that. Let's open this up. But the only problem here is that we don't actually get the spaceship in the background like we asked it for. So, we originally said change this to an image of the man dressed as an astronaut stepping out of a UFO spaceship. And it kept the plane the same in the background, but at least the image quality is better. That's pretty cool to be fair. So, let's open this up and see what we got here. So, we have a car made of small cars. That is a pretty cool image to be fair. The other cool thing about this is obviously it is open source, but also there doesn't seem to be a limit, right? So, you can create as many images as you want here. That is a usable image. I wouldn't I still wouldn't say it's as good quality as chat GBT40, but it's still decent. If you go back to the same prompt side by side, so chat ch versus bagel, you can see the image that comes back from chatbt by default is not as realistic, right? So this is a lot more photorealistic. It does have the spaceship with the astronaut, which is what we asked, but this one looks a lot more photorealistic to be honest. Let's now try something else. So we're going to say, okay, change this. Let's go back to this chat here. We'll say okay change this to a 3D animated image and we'll try the same inside chat GPT. So you can see for speed chat GPT is just winning every time but it's not very good at like creating those animated images. So for example we said change this to like a 3D animated image. Came back to us with this which doesn't look great. Just looks the same to be honest with you. Let's see what bagel comes back to us with. And if you're wondering like do you need an API key or anything to use this? You don't. You can just go to the demo. bagelai. org and then from there you can actually get access and you don't even need to log in. It seems pretty simple and easy to use. So no API required. So let's have a look at these side by side. It looks super weird. So if we compare them again, it's that image quality that's just not quite there. That's the only issue here is I can see the potential but it looks super weird. Which one are you going to choose? Even if this isn't exactly what you want, you're going to go with chat GPT, right? For sure. As well, what I would say here is this is decent, but if you tested it out versus something like Flux, you're going to get much better responses from Flux, which can also generate images. If

you're not aware what Flux is, let me show you an example. So, you can go into Flux here. You can paste in the prompt. You can upload the same image style that you had before like this for example and then we can compare all three outputs. Right? So if we have chat GPT over here, we got Flux in the middle and then we have bagel on the left hand side. If you were comparing them side by side here, let's have a look what Flux has done. Flux hasn't done anything to that image. All right, let's try the same prompt here. So we're going to go with change this image so it's helicopter inside flux. There doesn't I don't think there is an API available for bagel. I think it's just available locally. So you can host it locally with GitHub or you can also use it directly inside the chat. So here we have the images back from Flux. Let's compare these versus the helicopter one. So which one wins? I'm going to say Chach's image probably comes in second there. It's not quite as realistic as Flux is, right? If you look at Flux, they've added like the helicopter style to the plane over there. And then if we look at Bagel, it still comes in last just because the image is like it's quite pixelated, low quality, low res, etc. And it takes a lot longer, right? So, let's take this to another level. Now, we're going to say, okay, change this image so it's like a 3D cartoon style. We'll do the same inside chat GPT. Compare them side by side. So this is the cartoon style from Flux which I would still it's easy to beat Bagel at this point to be honest. If you compare them side by side, this is still not perfect but it looks better than Bagel. Let's have a look what chat comes back to us with. So if we compare these side by side, Chat GPT has easily won that, right? Like the image there from Chad Chep is way better. Let's pull this up. That's like proper cartoon style. Fluxes is a bit of a joke to be honest. Let's open this up. That is the image from sorry from Bagel. It's not even comparable to chatb. Look at those side by side. And then finally, we have Flux, which is okay. I wouldn't say it's style that we're looking for exactly, but it can do its magic. So overall, I'm

actually going to say that Bagel probably performing the worst out of all three of these. Chat GPT still comes in first. Flux is coming in second. Flux you can use for free though, which is good. Chat GPT obviously has limits even if you're on the pro plan and you pay in per month. And then Bagel has no limits, but it's a lot slower. However, it is open source and it is free. So if you

want to get all the video notes from today, feel free to get that inside the AI success lab. This comes with all the resources from today along with a bunch of courses and you can get that link in the comments description along with 100 different prompts for using bagel AI. Is it the GPT40 killer? I wouldn't say so at this point. I think it still needs a lot of work. However, the fact that it's open source and it can generate images for free, you can't really complain. And if you want to join this community, it comes with 9,000 members as you can see inside here. Additionally, if you want to get coaching support, feel free to get that inside the AI profit boardroom. Inside here, we make personalized video for you each week. We were actually running three coaching calls per week as well on AI automation. You can connect with me personally. Plus, inside here, we host all our best courses, workflows, AI agents, and every call gets recorded if you want to watch them back. I recommend check it out now. Link in the comments description. Appreciate you watching. And additionally, if you want us to implement all of the AI automation workflows that I show you, then we can just do that for you. Feel free to book in a call via this link right here, and we can book in a call, run through anything you want to automate, and then go from there. Just watching.