# Nano Banana Pro VS ChatGPT VS Midjourney VS Flux - Best AI Image Model

## Метаданные

- **Канал:** Skill Leap AI
- **YouTube:** https://www.youtube.com/watch?v=m-wH-VCYO_k
- **Дата:** 30.11.2025
- **Длительность:** 24:23
- **Просмотры:** 14,842

## Описание

Get all cutting-edge AI image and video models in one place:
https://artlist.io/artlist70446/artlist_aid=SkillLeapAI_3981&utm_source=affiliate_p&utm_medium=SkillLeapAI_3981&utm_campaign=SkillLeapAI_3981

I go step by step to show how I tested the top AI image models to see which one is the best right now. I tested Nano Banana Pro, ChatGPT’s image model, Flux 2 Pro, and MidJourney. I compared them using the same 15 prompts and judged how well each model handled realism, text accuracy, style, character consistency, image editing, hands, lighting, and more. I used the same prompts across all platforms and picked a winner in each test.

This helps me figure out which AI model is best for tasks like portraits, product photos, and thumbnails. I also explain which tools I used: Gemini, ChatGPT, Artlist, and MidJourney, and why I like using Artlist to test everything in one place. At the end, I tally up the results and share which model came out on top, plus where each one did best. I also link to a free prompt library you can copy and paste to test them yourself.

I put all my prompts in one page to make it easier to copy and paste: https://skillleap.futurepedia.io/pages/image-prompt-library/?utm_source=youtube&utm_medium=skillleap&utm_campaign=25

## Содержание

### [0:00](https://www.youtube.com/watch?v=m-wH-VCYO_k) Segment 1 (00:00 - 05:00)

I put the top AI image models to the ultimate head-to-head test to see who comes out on top. Just in the last week, we have a brand new model from Google called Nano Banana Pro, which is unbelievable. I covered it in my last video. We have another image model from Flux called Flux 2 Pro, which is also amazing. I'm going to show you that one, too. And we'll also use the chat GPT image model and MidJourney as well. I'm going to test them for realism, for text, for character consistency, image editing, and it's going to be part of a 15p prompt test and I'll pick a winner for each category and then we'll add everything up at the end and pick one winner for this video. So, Nano Banana Pro, I'm going to use that inside of the Gemini website at gemini. google. com. Chat GPT's image model, I'm going to use on the chat GPT website. I'm going to also test out Flux 2 Pro, which just came out. I'm going to use that on a different website called artlist and midjourney is at midjourney. com. So for the very first prompt, I'll just kind of show you how to make one in each platform and then after that I'll just kind of show you the result and then I have a resource where I have all the prompts that you could easily copy and paste that I'll show you in a little bit. Right now the prompt is going to be for realistic portrait test. Ultra realistic portrait of a woman. Soft window lighting, natural skin texture, sharp eyes with visible reflection, subtle freckle, smile, and shot on an 85 millimeter lens with shallow depth of field. Now, I worked as a cinematographer for a very long time, so hopefully I could be a good judge of these. And to use Nanabanana Pro, you just have to make sure you're on the thinking model here. And I choose the image model here from the tool dropdown, create image. And I'm going to set this up for chat GPT. I'll press the plus sign. Create image. Same prompt. Now with art list, this is the best place I found for actually using Flux. Flux is open source, so there's plenty of places you can use it. They don't really have their own dedicated website, but I'll use it here. Flux 2 Pro. And you can also choose your aspect ratio here. So I'm going to choose 16x9. And you could generate multiple images in this platform. Again, I'll just do one since the other two platforms only let you do one. And I'll go ahead and generate that. And for midjourney, I go to the create tab, paste that same prompt over here. Now, midjourney also has different settings here, but in this case, I'm going to just reset it to the default. And the model is just going to be the standard model. And we'll use version 7. Okay. Here's the result from Gemini. Really great. The reflection in the eyes is there. Subtle freckles. I mean, it's almost impossible to tell this is not a real photograph. Okay. Chip also followed the prompt pretty well here. This looks a little bit more AI to me though compared to what we got out of Gemini. This is out of flux. Now the one thing I could tell you is 85 mm usually will give you really shallow focus like this. So I think this actually followed that part of prompt pretty well. Although the person it created here looks a little bit too AI for me in the flux model. Okay, this is out of midjourney and just at first glance it did follow my prompt. The lens selection is also pretty good, but it zoomed it in so much that I really can't see that subtle background that the other ones had where I think it made it look more natural. This is almost too close to me. This looks a little bit like a video game character than a natural realistic portrait. Now, all four are on screen here, and I would pick the winner here to be Nano Banana Pro. It just looks the most natural to me. He even kept the window in the shot which really gives that direction of where the light came from. I really like that in these type of portraits. So that's going to be the winner. Obviously these are going to be subjective because again these are based on my taste in these. So you could pick your own winner, but I think this one is pretty clear to me. Our second test is a product photo test. So for this one, high-end studio photo of a matte black wireless headphone set on a glossy glass. So, this is going to be pretty tricky here. I explained the lighting here, clean shadows, and I also want to explain aspect ratio because right here inside of Gemini, it will follow if you tell it in the text prompt. It doesn't have an option that we had in some of the other platforms, but you can explain that. So, one:1 is going to give us a square image. So, I'll do the same with chat GPT. And for Flux 2 Pro in art list, I picked a square drop down here. And same with midjourney. Okay, Gemini followed my prompt pretty well, except you could see the light. Just I did not tell it not to see the light. I explained softbox lighting on the left. So, it decided to put that literally in the frame. So, it actually followed my prompt exactly right. But in this case, I could do a follow-up and make sure it crops that out because if I use this in any kind of ad, I obviously don't want to see the source where the light comes from. Now, chatd, I think actually got it right. I don't see the light here. I think the blacks are far too black though. These reflections are just kind

### [5:00](https://www.youtube.com/watch?v=m-wH-VCYO_k&t=300s) Segment 2 (05:00 - 10:00)

of getting hidden here. It just kind of goes to pitch black. So, I would have liked a little bit more detail down here. Now, this is flux. I think this one actually turned out really, really good. You see the details are still not getting lost in the black. The reflection is really nice. And I don't see a light source. I think this one's pretty ready to go here. Now, Midourney, I think, also did a really good job here with the reflection. The blacks are not too black. And this one's really close between flux and midjourney, but I think the midjourney version is ready to go. Okay, I generated this one to test character consistency. So, we're going to take this same image. This was created with a text prompt, but we want to put this exact same person in different scenes. So, I'm going to download this one from Gemini and then I'll upload it with another prompt to make sure it keeps the character consistent as we put them in different scenes. Now, here's what we got out of Gemini original creation here. This is in the coffee shop. Again, I would say that is definitely the same person. And he kept the clothing exactly the same. So, that's what we were looking for. Here is on the beach leaning on the rails. Very good there. And here it is in Time Square. Now, for Chant GPT original image, I'm not testing the very first prompt. This is for character consistency. But if I was testing that, this would be far worse than what we got out of Gemini. This definitely looks AI generated to me, this person, where the Gemini version did not. Now, in the coffee shop, it's okay. Like, it definitely changed slightly the way this person's eye looks, but it did keep the jacket and the shirt the same. Let's go over here. Yeah, it's kind of not great. This definitely looks a little bit slightly different. And yeah, this is almost entirely a different person now. But it's not horribly off. It's definitely getting a lot better. But if you compare it to Gemini, this definitely does not keep up. Okay, here are the four options I got out of Flux 2 Pro. I think it's really close. And the interesting thing is you see the subtle issues that he had with the jacket, the tears right up here. It definitely carried those throughout the scene here. Yeah, this is kind of a pass for me. Now doing the same thing in mid journey is a little bit more involved. So you got to go ahead and download the image and then upload it. But you have to choose this option first omni reference and then click or upload the image so it uses that then type in your text prompt then send this out. So it does require a lot more steps here to get it to actually use that reference. Otherwise the style reference is going to totally throw it off or the image reference. Okay. So these are the default settings. These are what I got out of Midjourney trying to do exactly the same thing I got out of the other platforms for character consistency test. Clearly Gemini or Nano Banana Pro here is the winner and I think Flux is in a close second place. Now let's see if we could generate an image and how well that image turns into a video. And just to keep it simple, we'll use the same video model which is going to be Veo 3. 1. So we'll generate the image first inside of Gemini. Okay. And I've uploaded the image on a new chat. And we're going to use Veo 3. 1, which is again part of a paid upgrade. So, you're not going to have this in any of the free accounts here. I'm going to describe what I want the astronaut to do. Slow steps forward. Looking around gently, right? A little bit of detail here. Now, Chat GPT doesn't have a video model right here. I can't turn it into a video. They do have one called Sora 2. That is the video model from the same company. But again, I want to use the same video model because this is a image test. So, we'll use Veo 3. 1 for that. Now, I'm going to do that in Art List 2. AI video tab actually has all the top AI models. So, we're going to use Veo 3. 1. And I'm going to upload that image. And 16x9. Let's do 6 seconds 1080p. And let's add audio here. And I'll add the prompt. Okay, that was a little bit hard to judge because I'm not judging the video model. It's the same video model in all these. I'm judging how well that composition turned into a video. And in this case, I think MidJourney actually wins. It's like the most cinematic and interesting looking shot out of the four. Now, I've been using Art List throughout this video, and it's actually my preferred way to use the top video models and top image models. And they're actually today's sponsor, too, for this video. And this platform I've used for many years, well before AI came out. So, my background, as I mentioned, is in video production. So, I've used Art List for all kinds of different creative assets that I needed for my client work. But with AI, Art List now gives you access to all the top image and the top video models in one place. As I've been showing you the AI image models here for

### [10:00](https://www.youtube.com/watch?v=m-wH-VCYO_k&t=600s) Segment 3 (10:00 - 15:00)

text to image, we have the best ones available here, including Nano Banana Pro and I prefer to actually use it here. Makes it really easy to generate multiple images at the same time to choose your aspect ratio here. And then you also have image to image, which in this case you'll see some other models that I did not cover in this video, which are also the top models, especially for image editing, like this one here. You see the flux models, you have Nano Banana Pro, the regular one. And literally the day these models come out, they added to this platform. Same with the video models. So if you want to use the top video models that I've been showing you here right now, Veo 3. 1 and Sora are kind of a tie, I think, between the two. Cling 2. 5 just came out though. This one I'm testing out to see if it beats the other ones. You could see you could use all of them here and generate again based on simple prompts here. And on top of that, they also have this creative assets library. So any sound effects you want, you don't have to generate it. They have a whole library. stock footage library here that you could get all kinds of different things from that you don't have to generate. A lot of times these are just easier to get this way. They have templates. They have lots all kinds of different things that you could get right in this platform without jumping between a bunch of different platforms and bunch of different apps. And the licensing is really straightforward. So anything you download here, including music, for example, is covered. So even if you do this for client and commercial work, it's all covered. And everything in this platform is really high quality and everything is handpicked including the best AI models. And this is really the only platform that gives you the top AI image and video models along with a catalog of worldclass creative assets like music, SFX, footage, pretty much anything you need for your entire project. So I'll put a link to Art List in the description so you can try it for yourself for free. For the next test, I want to see how well it does with hands. For some reason, AI for a very long time could not create hands that had five fingers. Let's see if that problem has gone away and how good these models could get that. Okay, Gemini. Yeah, that's a pass. We got five fingers. No issues there. Chat GPT, five fingers, but there's a little bit of issue with the thumb here. Has it like a weird split? Okay, Flux, five fingers. Overall, pretty good. Although this looks like an engagement or wedding ring here, but on the wrong hand. But again, it's a pass. It got the fingers really nice and realistic. The face looks good. Midjourney also really good. This is a tossup, but I'm going to give it to Nana Banana Pro. I think it looks the most natural. Okay, now let's see how well it handles text. So, I gave it a prompt here for a whiteboard, and I just kind of made up some text here on what it should put on the whiteboard, but quite a bit of text. Let's see how well it does there. Here's what we got out of Gemini. I've looked at it for 20 seconds and I cannot find a typo. This followed my prompt pretty much exactly right. I also wanted drawing descriptions. Pretty good there, too. Okay, here's what we got out of chat GPT. It was almost perfect. I found a couple of typos. For example, this word right here. It actually the cauldron here, it's spelled this way. So, chat GPT did misspell that from my text prompt. Now, here's what we got out of Flux. Now, Flux actually did not put all the text there. So, definitely not a pass. And even the text that it created is not very readable here. Now, this is what Midjourney did. So, I have no idea what's going on here. I tried it like three different times and yeah, I don't know what's going on. I can't read obviously any of it. Now, this was a close one. And even though Chhat GPT's layout is really good, it did have typo. So, I'm going to give this one to Gemini or Nana Banana Pro, too. Okay, our next test is style transfer test and I want to see if it could follow a prompt to do it in Pixar style. Portrait of a cat wearing a red scarf Pixar style. Okay, results out of Gemini. Definitely cartoony, but definitely not Pixar style. This is out of chat. Definitely not Pixar style, even though it is a cartoon drawing. Flux Pro 2. Yeah, this is definitely a lot closer to a Pixar style. And I have no idea what's going on with MidJourney. I mean, not Pixar style. So, this one had a clear winner and that is going to be Flux 2 Pro. Now, let's see how it handles a pretty complex scene. A crowded futuristic marketplace filled with people and robots. And I also added cinematic and realistic at the end. And this is the result out of Gemini. It did follow the prompt except the realistic part. This definitely looks like a drawing. It doesn't look like a real picture here. Okay. I'm not sure what chat GPD did here, but it's definitely not realistic. And yeah, it's really strange here. They get the neon signs right, though. Okay, flux. Not bad. In this case, you got all the text right, too. It looks like repairs, cyber noodles. Yeah, you really have to really dig in to see the issues. I mean, some faces are morphing a little bit in the distance, but pretty [clears throat] good. Okay, MidJourney actually created

### [15:00](https://www.youtube.com/watch?v=m-wH-VCYO_k&t=900s) Segment 4 (15:00 - 20:00)

an interesting world. I do like this setting. I like the camera angle, but the details really fall apart here. And if you look at the text, that's completely just gibberish here. Now, clearly Flux is the winner here. I think Gemini is in second place for this one. Now, here's an important test. Prompt understanding. So, this has a lot of little details like this elderly man, wooden bench, reading a book, and the book is blue, small white dog, right? Red umbrella. So, I wanted to have all these little details and how well it's going to follow that. Okay, let's see what we got out of Gemini. So, wooden bench, elderly man, blue book, red umbrella, white dog, misty fog here. The book also is called story for a foggy morning. That's not a great touch here, but that was not part of the test of the prompt. I was just trying to see how well he follows the other details of it. And yeah, it's really good. Okay, chat GPT. All the details are followed exactly the same too. No title on the book, which is actually good. Overall, it does look a lot more AI generated than the other one though. But for this very specific test, it followed every part of that prompt. Okay, Flux. Wow, that is really good. Look at the details on this book and the person. I mean, that one nailed it. Okay, midjourney again. I'm not sure what happened here. blue book, white dog, but two red umbrellas and the style is kind of strange. Okay, let me try one more in RAW. Okay, it followed the prompt this time, but also still does not look like a real picture. And just to show you, by the way, I don't have any of these settings changed. I just reset everything to the default setting. In this case, I think Flux again is the clear winner with Gemini a close second. Okay, now let's do a camera and lens test. action shot of a runner and it's a 35 millimeter lens and slight motion blur in the background. Sharp focus on the runner. So, this shows you how well it mimics what a real camera would create. Okay, Gemini. It looks pretty good. It's not very cinematic, but it definitely looks realistic and I think it followed the direction. The lens size looks good. The motion blur in the background, you could see motion blur here, but the runner is sharp. Okay, chat GPT. Yeah, definitely followed the directions for the most part, but I would just say this is really looking AI generated here. Okay, Flux. Yeah, that definitely looks good. I don't have any issues with this except the person is running in the middle of the street. And here is midjourney. And it's just far too much here. The motion blur just looks totally unrealistic. And it's even so much that it's affecting the runner. This one's a little bit tricky because I don't love any single one of these images, but I'm going to give this one to Gemini or Nano Banana Pro. Okay, for the next one, I'm testing creativity. I specifically want to see how well it follows kind of a weird prompt. A giant floating banana wearing sunglasses, hovering over a city skyline, people looking up. Okay, Gemini did follow the prompt. A lot of the people look clearly AI generated and their face did not get rendered, but again, that's hard when it has to create a lot of people. Okay, chat I guess followed the prompt, but again only five people. I didn't say specifically how many people. I'm not sure if I like the style. Okay, Flux also followed the prompt pretty good except ton of issues here. Like weird hands going on here. This person is just taking a picture of the wrong thing. Okay, this is Mid Journey. Again, kind of weird cuz it's just looks like a real banana. So, it didn't really fit the aesthetic of the rest of the image. This one's really hard, too, cuz I feel like there's no real winner here. But again, subjectively, I think I'm going to give this one to Nana Banana Pro, too. Okay, for the next one, let's actually see how well it follows lighting prompts. So, I specifically said there's a glass of water lit by a single candle light and softwarm light, deep shadows. Gemini. Okay, that's really good, right? There is no light coming from the left. Only direction of the light is here. is looking exactly right on the glass here. The reflection look good. This looks like a real photo. Okay, same with chat GPT. I do like the direction of light here. It's just overall though it does look AI generated. There's just something not quite right with it, but it did follow the lighting direction. Okay, Flux also followed the lighting direction here. And I do like how the background really falls into black cuz usually a candle wouldn't have that much throw here if it's the only thing in the room. And midjourney clearly did not follow directions here. There's clearly another source of light like a big window light coming through and candle is not the thing lighting the glass here. So definitely a fail for midjourney. Okay, this one is a close one again between Gemini and Flux, but I'm going to give it to Gemini too here. Okay. What about architecture detail?

### [20:00](https://www.youtube.com/watch?v=m-wH-VCYO_k&t=1200s) Segment 5 (20:00 - 24:00)

Wide shot of a modern concrete house on a cliff overlooking an ocean. Okay. This is from Gemini. He actually decided to create a new aspect ratio that we have not seen. Again, I'm not describing the aspect ratio, but it's wider than 16 by9. Little weird detail here though. What's going on with this? We're supposed to dive off the house. That kind of doesn't make sense there. Okay. Chat GPT. H. It's not bad, but it kind of looks totally like a fake house, right? There's no I mean, what's going on here? Okay, flux is really great. Directional light is good. Actual house, living room is filled with furniture. Okay, midjourney. Wow, this turned out really good. It's definitely hanging off the cliff quite a bit here, but that is the direction I give it. This one's a tossup between flux and midjourney. I'm not sure. I think I'll give this one to Midjourney. Now, for this next one, let's test out macro close-ups. So, this is going to be a honeybee on a flower. Okay, Gemini, really great. Okay, Chanty PD also did a pretty good job here. This is from Flux. Looks great. And Midjourney also looks great. Wow, this one is a it's a tough one. I think I'm going to give this one to Flux only because I like the details of the flower, too. That's nice. Now, for this one, I want to see how it could actually edit an existing image. So, this is one I actually made myself in Photoshop. And I'm going to ask them here to change the text and keep everything else the same. Let's see if it changes me. the text. Follows the prompts. This is what we got out of Gemini or Nana Banana. Definitely incorrect. It did not change all the text. It kept the Gemini part. So, that is not what I was looking for. Okay. Chat GPT. Well, first of all, it got the aspect ratio incorrect. So far, as I'm recording this, Gemini with Nano Banana gets the aspect ratio right only with Nano Banana Pro and Flux inside of Art List gets it right and MidJourney gets it right because they have those dials set for those. Chad GPT still can't get aspect ratio right. So, if you are trying to create an image in this size, which is a thumbnail size, it's going to fail. Okay, Flux best AI image model ultimate test. Text is right. Aspect ratio is right. Little Gemini logo is left over, but again, that is a little bit of a detail I did not cover in my prompt. I give it a pretty poor prompt just to kind of see what it would do. The text selection I'm not a fan of. This is not a real thumbnail I would use. It would need to be much better text, but I mean, it did follow my prompt exactly. And I have no idea what's going on with MidJourney. definitely recreated me here. The text doesn't make sense. And just to show you, I did this multiple times. Another one. Uh what I mean, I really don't know what's going on with Min Journey's image reference here. It just kind of followed the overall layout, but it changed every I mean, this is absolutely not usable in any way. Okay, this one clearly Flux wins, but Nano Banana Pro does do this very well if you give it the right reference and really spell out your prompt. Okay, let's tally everything up. And it looks like Nano Banana Pro is the winner. But it's really interesting because you could see in some use cases and some specific prompts, different image models do win out. So, it does still make sense to jump between at least multiple different image models and use the best one depending on your use case or test them against each other. And on platforms like Art List, for example, you saw you could just choose bunch of the different models with the same prompt in the same platform. So, you could easily compare them there and you don't have to have different subscriptions for different models. And I'll put a link to our website here. So, I've created this for our members, but I unlocked it right now. So, anyone watching this video could jump in and actually use this. But this has all kinds of different prompts in different categories that I've mentioned. You know, you could just go through this and copy any of these prompts and use them in your favorite image platform. And I have over I think a hundred so far, but I'll be adding more, including some of the prompts I used in this video here, too. So, I'll link this in the description along with everything else. And I also made a video showing you all kinds of different things. I think 25 different things you could do with Nana Banana Pro. So, I'll put that video here, too. Thanks for watching this one. and I'll see you next time.

---
*Источник: https://ekstraktznaniy.ru/video/11988*