What AI Image Generator Should YOU Be Using??

There are so many AI image generators out there now, and it can be really difficult to decipher which one to use. For very specific use cases, you've got Midjourney, which a lot of people claim is the best one out there. You've got Dall E3, which people are starting to call the Mid Journey Killer. You've got the new version of Firefly, Firefly Image two, which people are starting to say is just as good as Midjourney. And Dall E, you've got Stable Diffusion xl, which anytime I make a video about Midjourney or Firefly, there's always comments about people going, oh, why don't you just use Stable Diffusion? It's so much more customizable. And now we've got Google, who has their own image generator built into their generative search experience. And we've also got Ideogram, which was the talk of the AI art world about a month ago, when it was actually generating pretty decent text inside of images. So in this video, I want to figure out once and for all what is the best tool for very specific use cases. So we're going to go through, and we're going to do it at a rapid pace, and we're going to look at every single tool and we're going to grade it on accuracy, creativity, realism, illustration, logos and vectors, textures and background using text and images, the censorship in the images, the usability of the various user interfaces, and finally, the pricing of each one. And to wrap the whole video up, we'll talk about the pros and cons of each, and we should have a fairly definitive answer on which tool to use for very specific use cases with your AI generated art, let's go ahead and get right into it. The first thing we're going to test is accuracy. So for accuracy, we're going to test prompt adherence. We're going to pick a couple specific prompts. We're going feed that prompt into every single one of these tools here, and then we'll grade it on how accurately it followed along to that prompt. So let's go ahead and start with Mid Journey. And for our first prompt, we're going to use a photo of a green bus floating in space. Taking a look at these. I mean, it did a pretty good job. The buses are a little funky. This one's missing a wheel. This one seems to have some extra wheels on it. This one's not even floating, but pretty close to what I asked for. Let's give it a much more complicated prompt. Now, for our second prompt, we're going to use a sitting artist with a bucket hat painting a canvas of a three headed monster. And if we take a look at these, I mean, they're cool images. We've got the sitting artist. None of them are wearing a bucket hat. Only one of them appears to be painting on canvas. And I guess maybe you can argue this is a three headed monster if you look close enough, but not super close to what I asked for. For accuracy, let's give mid journey a 5 out of 10. Now, I don't feel like we can really use Mid Journey without using the Raw style in Mid journey because oftentimes if we use Style Raw, it actually adhere to the prompt a lot closer. So if we use our first prompt here, a photo of a green bus floating in space, then add dash, dash, Style Raw to the end here, theoretically it should adhere a little bit closer to the prompt and put a little bit less of Mid Journey's own style on it. And once again, it followed the basic prompt pretty well. This buses are a little bit funky, but they're definitely all green buses in space. Now let's do Style Raw for our second prompt. And once again, they're not perfect. None of these look like a three headed monster, but they all are a sitting artist, one of them, arguably two of them have a bucket hat. And it did catch that they're painting on canvas this time, so it got more of the elements. Right. Admittedly, this is a pretty complex prompt, so a lot of these tools aren't going to do this one great. But I have to give ROM mode a slight edge above the regular Mid journey because it did get a slight bit closer. So let's go ahead and call this one a 5.5. All right, now let's move on to Dolly inside of Chat GPT. So if I have Dolly selected here, I'll just go ahead and paste my prompt for our green bus floating in space. And if we look at these, these are all pretty dang good. This one here kind of made a yellow and green bus. But I mean, these other ones are all pretty great. Now let's give it our more complicated prompt and see how well it adheres to that one. Honestly, I'm pretty blown away. We have a sitting artist wearing a bucket hat painting a three headed monster on canvas. I don't think it could nail this better than did. So for Dall E3 instead of Chat GPT, I'm going to give this one a 9 out of 10 because it's nailing it. But Dall E3 is also available inside of Bing's image creator. And a lot of people have pointed out that there are differences between the images you get inside of Chat GPT and the images you get inside of Bing's image creator. A lot of these differences are going to become more apparent as we get into some of the later criteria that we're going to discuss. But let's go ahead and test these prompts inside of Bing's image creator here. Now, inside of Bing image creator, it did the green bus floating in space pretty well for all of them. No complaints here. And once again, Dall E3 pretty much nailed it with this one as well. We've got an artist who is sitting down wearing a bucket hat, painting on canvas of a three headed monster. All of these images are great. Going to give this one a9 as well. Dall E3 in both platforms pretty much nailed the accuracy. Now I want to test stable diffusion and my favorite platform to create with stable diffusion and is actually Leonardo. Now Leonardo has their own pipeline that actually does some additional creative effects to the images and it does use SDXL as part of that pipeline. So to compare the other image generators against sdxl, we're going to use Leonardo and its additional pipeline to sort of get the best that we can out of sdxl. And there we go. Two green buses floating in space. Both look pretty good. No complaints on either one of those. Now let's add our slightly more complicated prompt. And it didn't quite nail it, but in my opinion it did actually do better than mid journey at adhering to the prompt. We've got our sitting artist. We've got a bucket hat. They both have monsters in it. Neither of them are three headed monsters. Let's go. 6.5. All right, let's move on to Firefly Image 2, the newest image generator in town. And here are green buses floating in space. It gives it kind of more of a cartoony style, but they are all green buses floating in space. Now when we give it our second prompt here, they aren't bad. All of the artists are sitting. One of them is wearing a bucket hat. Ish. They're all painting on canvas. They're all painting monsters, but none of them actually are painting the three headed monster. So it sort of missed that one little element of the three headed monster. I'd say it's on par with what we got out of Leonardo. So let's go ahead and give it a 6.5 as well. All right, now we have Google's generative search experience. And they just recently rolled out the ability to generate images directly from search. If I type in my Google search here, create and then paste in our prompt prompt a photo of a green bus floating in space. You can see it now generates images up here. And here are our green buses floating in space. Now it seemed to take some artistic license and turn it into like a VW bus, but I'll take it. It followed the prompt. Now if I give it the prompt, create an image of a sitting artist with a bucket hat painting a canvas of a three headed monster. I actually get this. AI generated images aren't available for this search. Try creating something else. But asking it to go one more time. It actually generated the images the second time around. Artist sitting canvas. It looks like it danced around the three headed monster. This one seems to be a two headed monster and this one seems to be a four headed monster. But it definitely seems to be closer than Firefly in that it figured out that it needs to create a painting with a multi headed monster. So let's go 7.2. And finally we have Ideogram, which is a tool that a lot of people were talking about from for a very short amount of time. And here's what it came up with for us. We've got a sort of cartoon of a green bus in space. We've got a green bus definitely not in space. A green bus definitely not in space. And a green VW bus definitely not in space. Not amazing. Plugging in our more complicated prompt here. Let's see what we get out of that. And here's what we got from that. So this one got the sitting and maybe the bucket hat a little bit, this one actually got the three headed monster, the sitting in the bucket hat. But seeing as how it sort of messed up the really big basic one with the green bus floating in space and kind of did an okay job with the second, more complicated one, especially this one right here which is fairly close. I've got to put it at about the same level as Firefly and give it a 6.7. And looking at our first criteria of accuracy, here's how it all kind of shook out. Mid Journey got a 5. Mid Journey Raw 5.5. Dall E3 got a 9. It really, really nailed the prompt adherence. SDXL 6.5. Firefly 2 6.5. Google I gave a 7 point and ideogram a 6.7. And now that you have a fairly decent idea of how I'm going to be ranking these, I'm going to try to pick up the pace in this video and just crank through the rest of these prompts. I'm going to tell you the prompts that I'm going to use, show you the results, give you a score, and we're just going to keep this thing moving. So for creativity, we're going to give it barely any information in our prompt and make a very subjective judgment call around which of these tools are more creative and which ones just kind of make boring images if you don't give it much context. So jumping back to Mid Journey, I'm just going to give it the prompt. Beautiful creative epic RGB image. And here's what Midjourney generated. And I actually think all of these look really amazing. They've got a lot of color, they've got some amazing contrast. Interesting how with that same prompt, they all kind of have a person looking off into this colorfulness and then a random like eagle. But I would say as far as creativity goes, that's pretty solid. Now let's just give it a one word prompt. I'm going to give it the prompt. It clearly associates beauty with women. And all of these are pretty dang creative. I like all of them. They all came out really good. For Mid Journey, I gotta give creativity a high ranking. I'm gonna give mid journey a 9 on the creativity scale. The prompt, a beautiful creative epic RGB image in the style of Raw came up with these. They look amazing, but I'm not a fan of the fact that they tried to put some random letters in there. So I gotta dock it a quaint for that. But this tree image here I think is amazing. If you cropped out these letters here, that looks really cool. This image is really good here. Great image here. It looked like it tried to incorporate some letters. Same with this one. And here's what it generated. When I plugged in Beauty in Style Raw, we get a more realistic looking image for three out of four of them. Still pretty good. Definitely not as creative as the non Raw version. So let's go ahead and give midjourney raw an8. Using Dall E with ChatGPT. Here's our epic RGB images. They are all pretty dang creative, pretty dang colorful, but they sort of lack some of that contrast, some of that depth that we saw inside of Mid Journey. They are all very cool, but I have to give Mid Journey a slight edge and when entering the prompt beauty. These are the images we get. They're pretty good, but in my completely subjective opinion, I wouldn't say that they're quite as creative as what you get out of something like Mid Journey. So I got to give Dall E3 with ChatGPT 7.5. Now testing Dall E3 inside of Bing image creator, I feel like our Epic RGB image is even less creative. It just kind of made a whole bunch of colorful swirls, and when given the prompt Beauty, it kind of created three women's face that all look like they could be the same person. Interestingly, I'm finding Dall E3 instead of ChatGPT to be more creative than Dall E3 instead of Bing Image creator. And I think that might be because when you're using it in chat GPT, I think ChatGPT might be adding some additional context to the prompt to try to get better images. So for being image creator without much context to the image, it's not great. I got to give it about a six. Now jumping over to Leonardo, given our Epic RGB prompt, we get these two images and they look great. I mean, we're back to sort of Mid Journey level. I feel like Mid Journey might have a little bit deeper contrast and both of these images are very similar. I would think with such a broad prompt, we would actually get more unique, diverse images, but these are both pretty dang good. Now, given the prompt Beauty, once again, we get two images of females. Both of them seem to have a bunch of flowers around the hair. It's just interesting to me that I can give such a nondescript prompt and it would create multiple images that are essentially the same subject, where when we look at something like Mid Journey, especially the Raw, these are all kind of different from each other. I do, however, think this created some pretty good images. It's definitely better than what we saw from the creative images from Dall E. It's got more contrast, and to my eye, they're just more impressive looking images. I'd put it more on par with that Mid Journey Raw model and give it a 7.9. Using our beautiful creative Epic RGB image prompt inside of Firefly 2, we get a bunch of colorful swirl images. They're cool. They're definitely nowhere near the level of what we'd get from the creativity of something like Mid Journey or even Dall E3 and chat GPT. And when I enter the prompt Beauty into Adobe Firefly, I actually get a pop up that says the prompt is too short. And then I get pretty much similar images that almost look like they're golf course or a park with some trees and a lake. So from a creativity standpoint, so far, Firefly has been probably on the lowest end of the spectrum for adding their own unique creative spin. So for Adobe Firefly, I'm gonna give it a five. Using Google asking the same prompt, I actually get four very unique images. This is the first one where all four images are completely different than each other. However, I feel like it's really missing the colorfulness. Right. I asked for an RGB image and this image is really like only a few color tones. Same with this image. This one has some really colorful neons to it and this one does have a rainbow going through it, but it's really missing that ultra colorful element. And when asking it to create an image to represent beauty, it says an AI powered overview is not available. I've actually searched it a couple times now and it's refused to create this image a few times now. The first image was a little bit better than Firefly, but probably not quite as good as Dall E3. So I got to give it a 4.5.3. And here's what ideogram generated with our RGB image. Once again, four completely different images from each other. All pretty creative. Doesn't have that same depth of contrast that Mid Journey has, but they're all really cool, really creative images. I especially like this lion and this person walking inside the colorful woods. The prompt beauty. Once again we get four female images. This one has an extra finger, this one has some wonky fingers. They're not amazing images, but I did like the flat first RGB images that it gave me. So definitely better than Google and Firefly, but absolutely not on the same level as Mid Journey. This one I'd have to give 6.8. And here's where we stand on creativity. Mid Journey definitely wins out in the creativity department, followed closely by SDXL and Leonardo and Dall E3, the ChatGPT version. Now let's take a look at realism. This one we're going to use the prompt image of a couple holding hands in front of the Eiffel Tower. And I'm just going to use one prompt across the board for all of this because the people and the location are either going to look realistic or they're not. In order to keep this video under an hour long, we're just going to play around with the single prompt. So using Mid Journey we get some pretty dang realistic images. Upscaling this one here, I would say this one could probably fool some people other than the fact that these lights appear to just be floating in the sky here. Eiffel Tower Looks pretty good. The people look pretty good. Although there's not much detail in the face or anything like that. And he appears to be grabbing her arm weirdly. But I'd say on a realism scale, I'd give this image a 7.5. I think most people would look at this and know it's AI generated just by the floating lights in the sky here. Now, looking at midjourney style Raw, I actually think these images are even more realistic. If I upscale this one here, I have a hard time picking out anything about this one that screams AI. Now, I've actually never been to the Eiffel Tower in France, so this might not be what it looks like to the walk up of the Eiffel Tower at all. And if you know what that looks like, then you'd probably be able to spot whether or not this is not real. But these people look super real. The sort of slightly blurred out tower in the background looks realistic. The light posts look realistic. The people in the background look realistic. This image is pretty convincing. Let's go. 8.5. Now jumping into Dall? E 3. None of these really look super realistic. This almost has like a Pixar kind of quality to it where it looks computer generated. Same with this one. Especially with all the fingers going on here. Something looks off there. They just look slightly cartoony to me. I'll give it a five. Now plugging it into Bing's image creator. They actually look slightly more realistic. This image looks more realistic than any of the ones we just looked at inside of chat GPT's Dolly. This one, there's some weirdness going on with the face. You can definitely tell this one's AI generated. This one's looking pretty good too. But once again, you can see a lack of detail in the facial features here. And then this one's just not realistic at all. It looks almost like a hand drawing. But I would have to say that this did a better job at realism than Dall? E 3 inside of ChatGPT. I'll give it a six. Now let's take a look at what Leonardo generated. That's looking pretty realistic. Definitely better than what we were getting out of Dall? E. But still some weirdness in the face. You look at these faces and there's just something a little off. There's a lack of detail around the lips and the teeth and a lack of detail around the lips on this. And then when you look at the way they're holding hands. Yeah, there's kind of an issue with the hands on this One still. And then if you really look close, the Eiffel Tower has too many legs here. There's a leg here, a leg here, a leg here, a leg here, here, and then some more legs in the background. So, you know, I got to take away some points for that. But overall it's looking pretty decent. And then here's the second one that it generated. And if we look at the faces here, definitely a lack of realism in the faces here. And the Eiffel Tower almost looks like it's a tangle of cords or something. The structure definitely looks like it's bent or something right here. And there's just some funkiness going on with the tower. I'd put it on par with what we just saw out of Dall E3 with Bing's image creator. Maybe a slight step up. Let's give it a. A 6.3. Now, when I plugged it into Firefly image 2, it actually ignored the fact that I wanted it in front of the Eiffel Tower. So when I look at these images, they look fairly realistic until you look closely at the face. And the faces definitely have something funky going on in this one. This one looks pretty dang realistic actually. The hands look good. Maybe there's a little bit of a lack of detail in the face. Now, it didn't really follow my prompt instructions, but I'm judging criteria on realism and not accuracy of prompt. And this is looking pretty good. It looks like it could have been taken on just like a lower megapixel camera. This one's looking pretty good too, but there's something funky going on with their hands. I think she broke his wrist or something. This face is looking pretty realistic to me. This one, there's a little bit of funkiness going on around the teeth. And this one's looking pretty real as well. I'm going to try it one more time, see if it gets me the Eiffel Tower in there. One or more of the words may not meet user guidelines and were removed, but these ones are all looking pretty realistic. They just ignored my prompt. And people holding hands seem to be a problem, but as far as faces go, it's probably the most realistic we've seen. It's hard to put it on par with what we saw out of midjourney here. So I'll put it somewhere between the two mid journey versions and give this one an 8 for realism. Now, looking at what Google gave us, neither of these are even close to being realistic. Let's give Google a three on this. And finally, taking A peek at ideogram. Definitely some issues with this image. The woman looks like she's nine feet tall, or the man is like three feet tall. I don't know what's going on here. This face will give you nightmares. There's pretty much no detail in the people here on this one. And once again, I don't know what's going on with this person's face. I mean, the Eiffel Tower looks not too bad, but that's more because it's sort of blurred out. I'll give it a four. Here's where we stand on the realism scale, with Mid Journey Raw being the most realistic, Firefly 2 being the second best, and Mid Journey without using RAW being the third best. All the rest kind of left a lot to be desired. Now let's look at illustration. The last one was looking at how well it did realistic images. This one's going to look at how well it does sort of drawings and illustrated work. Now, for this test, instead of Midjourney, I'm gonna go ahead and switch it over to Niji Mode because we know this is more of an illustration style model. For this one, we're gonna use the prompt in Anime Girl with Braids in the neon streets of Tokyo. And here's what we got out of Mid Journey using the Niji Mode. I'd say all of them look pretty good. I love the contrast that we get out of Mid Journey. To this day, I feel like it does sort of colorful, contrasty images better than any of the other platforms. I'm gon go ahead and roll with an 8 on this one. Now, you can't use Niji Mode and Style Raw at the same time. They're not compatible. So when I use Style Raw, it actually tries to make a realistic image of an Anime Girl, which these images look great. But I'm trying to get an illustration and Style Raw doesn't really do that well. So for that reason, I'm going to give the Style Raw just like a brie. It's just not what the RAW mode is designed to do. Here's what Dall E3 inside a of ChatGPT created. They all look really good, really coherent. They don't have that same level of contrast that we get out of Mid Journey, but they're all still really solid images. I'm gonna give it a seven. Now looking at Bing Image creator, very similar style to what we got out of Chat GPT when looking closely at them. This one has some extra funky braids going on here. That one's pretty good. They're all decent images. I would say on par with what we got out of the Chat GPT version, but still not quite that same level of contrast that we get out of Mid Journey. So same as the ChatGPT version. I give it a Looking at Leonardo, here's what we get and Leonardo also has that contrast, that depth that you see in Mid Journey as well. Here's our first image. Here's the second image it generated. Both of them are really, really solid. I've got to give it a level above what we're seeing out of Dall E3, but maybe not quite as good as what we got out of Mid Journey. Maybe on par. I'm going to go 7.8. Here's what we get out of a Firefly Image 2. All of these are really good. They actually have those two deep dark contrast. They're all very, very solid. I mean, I can't complain about any one of these images. They're on par with what we've seen so far. It's hard to give this the same rating as what I gave this. I'm going to go 7.5 once again. Trying it on Google. It says AI generated images aren't available for this search. Try creating something else. I tried a whole bunch of prompts. I finally got it to create some images by typing create an illustration of an animated character and we actually got a spongebob, a Pokemon, the Little Mermaid and Mickey Mouse characters that are existing ip, which is fairly interesting. But it does do illustrations pretty well. Just it wouldn't adhere to the prompt that I gave it. Unfortunately. I give it the prompt created an illustration of an anime character. I also get some pretty solid illustrations. I mean that is a really good image. So prompt adherence not great or very, very censored. But the actual illustrations it gives is pretty good. I'm going to give it a 6.5 just because I had to test so many different iterations of this prompt to actually get something out of it. And then finally looking at Ideogram, these are all pretty good. I wish there was a little more detail in the face. That one looks good as well. Some weirdness going on the face here. Not bad, not bad. I'll give this one a 6.8. And here's what we got for illustration. All of them, to be honest, are pretty solid at creating illustrations. Other than trying to use Midjourney Raw, which is just not designed for midjourney, I think was the best. Leonardo and Firefly 2 are up there with it, but across the board. All of these tools are pretty decent at creating illustrations. And also keep in mind these numbers are ultra subjective. I'm literally just picking them at random based on my own subjectivity. You might score these totally different. All right, now we're going to test logos and vectors with each one of these. Let's see which one you'd want to use if you're trying to generate a logo. So for this one we're going to use the prompt simple flat vector image logo of a wolf on a white background in Mid Journey. Here's what we get. All of these look pretty solid to me. I especially like the top left and the bottom right one. I'll give it an 8 using Style Raw. Once again, all pretty good. I would say it's on par with the Non RAW version. Also an 8 looking at Dall E3 instead of chat GPT. I really like this one, how it's sort of squared off. This one leaves a little bit to be desired. So does this one. This one's not bad. They're pretty good. I think Mid Journey did a slightly better job, but they're pretty good. 7.5 using Bing Image Creator. Once again, all pretty good. I would say it's on par with what we got out of the ChatGPT version. Also 7.5. Now, looking at Leonardo here, they're a little less simple. I like what it generated, but it's not exactly what I was looking for when I asked for a simple flat logo. So while these images look great and they do look like vectors, they got the vector part great. They're not quite as simple as I'd like. If I was to generate a logo, I'm not sure that this would be the exact tool I would use. But also Leonardo has so many other models that are probably better for this. So I'd probably test different models to come to a logo using Leonardo. But for the sake of this demonstration, I'm going to give it a six. Dolby Firefly actually did a pretty good job. These all look pretty solid. I have to say that these are probably better than what we got out of something like Dall E. So I'm going to go ahead and say 7.8 using Google. It actually kind of nailed what I was asking for. These are all simple flat vector vector potential logo images. It did a really good job. I'd say an image like this is actually slightly more impressive than even what I got out of Mid Journey here. I'm gonna go 8.3 and then looking at Ideogram, it Did a pretty good job as well. I'd say these are on par with what we got out of Firefly. Fairly similar style to what you'd get from Firefly. So I'm going to match firefly and say 7.8. So taking a look at logos and vectors, all of them actually did a pretty good job. But surprisingly, Google was the standout for me as far as logos go. Follow mostly by Mid Journey and then Firefly 2 and Ideogram both did a great job. Dall? E did a great job. The only one that sort of didn't meet my expectations specifically for logo design was Leonardo. All right, now let's look at textures and backgrounds for this. We're going to see how well it can do with tiling and creating a sort of textured tile background. For this one, we're going to use the prompt colorful circuitry. And then inside of Midjourney, you got to give it a dash, dash, slash tile command to get it to actually do the tiling effect. So Mid Journey does a great job at creating these tileable images here. If you were to lay one of these four images side by side by side, they would all line up and match. It works with both regular Mid Journey as well as Style Raw, both very, very similar images. But I feel like this test is almost like a pass fail test. It either does tiling or it doesn't do tiling. And since both versions of Mid Journey do tiling, I'm going to go ahead and give them both a 10. Now, using DOL3 with chat GPT, I gave it the prompt colorful circuitry, but then I also gave the additional context make the image a repeatable tile. And you can see that it did exactly that. If I was to download this image here and set it as my desktop background, you can see that it doesn't actually tile that well. You can see all the seams where it's supposed to tile. Comparing that one to this one that we made in Mid Journey here, I set this one as the desktop background. You can see that it actually does seamlessly tile. You can. You can't find the seams on this one. So although Dall E claimed it made a tileable image, it's actually not. It sort of hallucinated that it's tileable. If we just look at some of these other images, you can see there's the yellow line going here, which means that yellow line should be connecting somewhere down here, but we're not seeing that yellow line pop in. So this one is not actually tiling. They're not doing what they said they would do. And since this is kind of a pass fail, I kind of got to give it a zero. Now, if I look at Bing Image Creator, I gave it colorful circuitry, repeatable tile, and I can tell just by looking at these that these are also not repeatable. If we take a look at this image, for example, you can see this blue line kind of comes up here and then goes off to the right, top right of the image, which means down here, we should see that blue line continued somewhere, but we're not seeing this continuation. So this one also would not properly tile. Also pass, fail. It's a zero. Now, if we look at Leonardo here, if I come over to my settings on the left side, you can actually see they have an option for tiling. If I flip this on, it gave me two images that are tileable. So let's go ahead and download one of them. I'll set it as my desktop background. And as you can see, it is properly tiling. And I can't tell where the seams are. So Leonardo pass. That's a 10 for Adobe Firefly. I gave it this prompt, colorful circuitry, repeatable tile, and these ones might tile. I'm having a hard time telling just by looking at them. Let's go ahead and pick this one here. Now, just looking at this one, it looks like it probably will tile. You can see how this line goes off here and then would connect right here. This line goes off here and would connect right here. Here. Same with this and this. Let's go ahead and download it. Set it as my desktop background, and if we take a look at it, it's definitely tiling. I'm not seeing where the seams are on this one. That one is a pass. It gets a 10. I can't get Google to create any sort of tile, so that's a zero. And looking at Ideogram, these do look tileable to me. Let's go ahead and try one. Setting it as my desktop background, it is attempting to tile, but you can definitely see the seams where it's not quite doing it right. There's a seam right along here and a seam right along. And the fact that we can see the seam, it didn't pass my tiling test. So ideogram. Sorry. So for textures and backgrounds, basically a test of tiling. Midjourney can do it. SDXL Leonardo can do it. Firefly 2 can do it. Dall E3, Google and Ideogram all seem to struggle with it. Still now let's see which ones can do text in image. This is a newer thing that's been happening in AI images lately where more of the models are actually allowing you to do this. This one I'm going to use the prompt a penguin holding a wooden sign that says subscrib to Matt Wolf. So taking a peek at Mid journey, we got the penguin holding a wooden sign, but it can't really figure out the text yet. This one also puts some human hands on the penguin. So mid journey got to give it a zero on this one because it really can't do text at all. Looking at Mid journey, raw, same thing. We got a penguin. This one even looks like a sort of realistic penguin. It's holding up a wooden sign, but the text is just not even close. Also a zero. All right, so looking at Dall E inside of Chat GPT, we got four penguins all holding up wood signs. Subscribe to Matt Wolf. This one nailed it. We typo and subscribe here. They also spelled my first name wrong. Subscribe to Matt Wolf. Once again, a typo in the word subscribe and also a typo in the word subscribe. So Dall E3 can do text and it does pretty good. I got one out of four on this. I'm going to go ahead and give dall E A 7.5 on this one because it can do text but there's room for improvement. And if you wanted to get text out of Dolly, probably just got to prompt it a couple times and eventually it will nail it. Perfect. Looking at Dall E inside of Bing's image creator. Subscribe to Matt Wolf. Subscribe to Matt Wolf. It missed an I here, but both of them kind of nailed the text. And just like the other version of Dall E that was in Chat GPT, it prompted a couple times. It's going to nail the text on one of them. 7.5. Also now taking a look at Leonardo, we've actually got some pretty good looking penguins holding up some wood signs and it's actually closer than I thought we would get. Right. So this penguin, you can tell the words aren't quite right. It spelled my last name right. My first name was wrong. Subscribe. And they put torte instead of two. This one is really only missing the C. And subscribe. It added an extra F here and then said my name was Matt Whitwill. So it's getting closer than midjourney, but it's still not really doing it great. I'm going to give it a 4.5. Looking at Firefly 2 here. It's not great. I would say slightly better than what we were getting out of Mid Journey, but not by much. The images look good. They got the penguin, they got the wood sign, but it's not even really coming close to the words. It's got some of the letters right, which is better than what we can say for Mid Journey, but still not great. I'm going to say 1.5. You're not going to go here if you need an image with accurate words in it. So in order to get this to work on Google, I had to change the prompt a little bit. I typed create an image of a penguin holding a wooden sign that says subscribe. If I put it in quotes, it didn't work. And if I put subscribe to Matt Wolf, it didn't work. But when I put this, you can see that there's a penguin holding a wooden sign that says subscribe. But if I say to to Matt Wolf and search it, it doesn't seem to work. If I add quotations around, it also doesn't seem to work. Nonetheless, I can get it to generate images with text, and the text is correct every single time. So it's working well. I just can't seem to get it to do more than one word for some reason. Maybe you know what I'm doing wrong. If you do, let me know in the comments. But because it's actually getting the words right and it's following the prompt I asked it to do, I'm gonna give this one like a 7.7. It would be higher if I. If it actually was able to do more than one word at a time for me. And now when I look at Ideogram, the one that sort of kickstarted this ability to add text and images, it's pretty solid as well. It didn't get the whole thing. This penguin just says subscribe. This penguin just says Matt Wolf. This one says subscribe to Matt Wolf. And this one says subscribe Matt Wolf. Other than the fact that it didn't get the entire statement onto the sign, there's no typos. And. And it followed the prompt pretty well. And just like Dall E, if I was to generate it again, it would eventually get the perfect image for me. So I have to put ideogram on the same level as like Dall E for the text inside the images. And I'm going to give it a 7.6, 0.1 more. So for text and image, Mid Journey gets a 0 on both accounts. Dall E3 gets a 7.5 on both accounts. Leonardo and Firefly not really doing it too well. And then Google and Ideogram also do fairly decent. So if you need to get text in your image, go to Dall E3 Google or Ideogram to get it. Now let's look at censorship and see if we can get it to generate images that use either celebrity faces or IP and logos and things like that. But for this one, I'm going to do Tom Hanks standing next to a Stormtrooper. So we've got Tom Hanks, a celebrity, and a Stormtrooper, which is obviously Disney ip, and let's see if it'll generate that. So looking at Mid Journey, it made both Tom Hanks and a Stormtrooper. Pretty surprised by that. I didn't actually think it would. And the Style Raw also did the same. They're pretty dang realistic, too. Now, given the prompt spongebob with Super Mario, it actually managed to pull in Mario decently well, but couldn't figure out the spongebob using Style Raw. Same thing. It got Mario pretty well, but really no spongebob there. So surprisingly, Midjourney isn't really censoring celebrity names like Tom Hanks or IP like Stormtroopers and Mario. But there are some celebrities that it just will not create. For example, if I give the prompt Joe Rogan behind the microphone, none of those really look like Joe Rogan, but surprisingly, it generated everything I wanted. I don't think it doesn't know what spongebob looks like. It's just not very good at doing two characters next to each other. For example, here's what it gives me when I give it the prompt Spongebob Squarepants. So I'm going to give both Mid Journey versions 7.5. Now, looking at Dolly3 instead of ChatGPT, it couldn't get Tom Hanks, but it did get the Stormtrooper. And it says, here are the images of the recognizable male actor standing next to a Stormtrooper. Now, if I do the prompt spongebob with Super Mario, it says I'm unable to generate images of spongebob with Super Mario due to content policy restrictions. So although it generated a Stormtrooper, it wouldn't generate Tom Hanks, it wouldn't generate Super Mario, it wouldn't generate spongebob. So I'm gonna give it like a three. When I jump into Bing Image Creator and put Tom Hanks standing next to a Stormtrooper, it says this prompt has been blocked. It flagged the prompt because it may conflict with our content policy, which is a little frustrating because it was letting me do prompts like this, like two days ago. So this is a brand new thing where it's no longer letting me create whatever I want. However, when I do SpongeBob SquarePants with Super Mario, it gives me both of those characters just fine. You can also see in a previous prompt, I was able to get Larry David with Serena Williams. So it does appear that it's got some censorship, but not as much as what the ChatGPT version has. So I've got to give this one like a six because it's still letting me generate these images right now. But I have a feeling a lot of this IP is going to be totally blocked in the near future. All right, looking at Leonardo, we've got Tom Hanks with a Stormtrooper. Pretty dang convincing too. This stormtrooper has a random hand on his shot shoulder, but other than that, they look pretty dang good. Using spongebob squarepants with Super Mario, it makes an attempt. They don't look great, but it's definitely not censoring them. As far as I know, Leonardo doesn't censor anything. You could make adult content with this one if you want. So Leonardo, as far as censorship goes.