Transcript for:
Nano Banana Image Model Overview

People are doing some absolutely wild things with the new Google image model that everybody's been calling nano banana. Things like the ability to do this or the ability to completely change the character that you're seeing on the screen like I'm doing right now. So, in this video, I want to give you a deep dive breakdown of all of the coolest ways I've seen people using Nano Banana. We've got like 50 different use cases of Nano Banana. This is going to be a rapidfire video where I just rattle off tons of different ways to use it. And the idea being that this gets your creative juices flowing and you really get a grasp of how powerful this model is and how many different things you can do with it. Now, before I get into it, I do want to share a quick resource that I think a lot of people will find helpful. Google itself actually put out a prompting guide to show you how to get the best results out of Nano Banana. I will make sure it's linked up below, but when you're on this page here, click on image editing. They do give you some prompts for image generation, but it's not good at image generation. It's amazing at image editing though. Scroll down and you'll see a whole bunch of prompts for generating images and we will play with some of those throughout this video. I've made several videos where I talk about Nano Banana. I don't want to get too deep into how to use it. My favorite place to use it right now is through Google's AI Studio, but you can also use it over on Gemini. You can use it in Higsfield. You can use it in Crea. You can use it in Freepic. You can use it in Adobe Firefly. Pretty much all the platforms have this new model in it, but I'm going to use Google AI Studio for these examples. It's really good at taking images of two different people and blending them together. Like this example of Billy Isish, Michael Jackson, and then a selfie of the two of them together. If I give it an image of myself and an image of Keano Reeves and say, "Put these two people together during a fancy gala taking a selfie together," I get this image. I don't know why it made us holding hands, but it did what I asked for. It's also really good at removing people from images if there's someone in the image that you don't want there. If I take this image that I took in Hawaii of all the sea turtles on the beach, but I want to get rid of all the people in the background, I can give it the prompt remove all the people in the background. Upload my image. I get an image back that's the exact same image, but with all of the people removed. It's really good at removing objects as well. Here's an image I took in Colorado recently, but I want to get rid of the traffic cone and the construction sign here. Google's prompting guide actually has instructions for this. So, I'm going to go ahead and use the prompt that they created here and use that on my image using the provided image of Colorado. Please remove the traffic cone and construction sign from the scene. Ensure the change is description of how the change should integrate. I'm just going to remove that. I think it will know to just remove these things. And we can run it. And we get an image back with those elements removed. Let's take a closer look. You can see everything's still in there, but it removed the cone and the little traffic sign. The rest of the image stayed exactly the same. It's great at replacing objects. Here's an image of me looking at my phone. If I tell it to replace the phone in the person's hand with a banana, well, we get me holding a banana. I can change the location of where the photo was taken. So, let's go ahead and copy this image. Paste it back down into our prompt box here and say, put this person holding a banana in a monkey enclosure at the zoo. And here I am holding a banana inside the monkey enclosure at the zoo. I can rotate the person in the image. Let's take our image from right here. Copy our image, paste the image back into our prompt box, and give it the prompt. Change the camera angle so that it's looking at this person from the side. And we get an image of myself looking at a banana from the side view. and it's super accurate. Maybe you need a professional headsh shot, but all you have is an image you took quickly on your computer. Toss that image into Nana Banana. Give it the prompt, make a professional studio quality headsh shot of this person and looking pretty sharp. Pretty pretty pretty good. Want a full body version of the image? Just tell it create a full body professional photo of this person. A few seconds later, it did just that. Now, it does lose quality a little bit, especially when like the head gets a little bit smaller, so you probably want to run it through an upscaler, but it's totally capable of doing what we asked it to. Maybe you want to see how you'd look in a specific outfit, like maybe this outfit could look cool on me. Upload an image of yourself and then upload an image of the outfit that you want to put on. Google gives us a very specific prompt that works well for this, so let's go ahead and use it. Create a new image by combining the elements from the provided images. Take the person from image one and give them the outfit from image two. The final image should be the person in the first image wearing the clothes of the person in the second. And just like that, I know what I would look like inside of this outfit. Want to try a different hairstyle? Upload an image of yourself. Upload an image of the hairstyle you want to try. Let's use our same prompt again and just tweak it a little bit to get it to add the hairstyle. We'll give it this prompt here. And well, it tried its best. You can also do the same thing by just giving the original image and giving it a text prompt of what you want, like give this person a colorful mohawk. And honestly, I actually think it kind of did a better job than when I combined the two images. You can have it change the color of objects on an image. Like, let's say I want to change the color of this green chair to purple. We'll pull in the image, give it the prompt, change the color of the green chair in the foreground to a purple chair. And without changing anything else in the image, we have a purple chair instead of a green chair. It can also colorize images. Here's a black and white image I found online. Toss it in. Give it the prompt. Colorize this image. 7.8 seconds later, we have a perfectly colorized version of this image. Let's take this exact image, paste it back into the prompt box, and say, make the colors more vibrant, colorful, and modern. 9 seconds later, we have an even more colorful version of the same image. You can also change the perspective on an image. Let's take our image here and give it the prompt, give me the view of this from above looking down on the two women. And 9 seconds later, we have this image from a different angle looking at these two women. And if we look back at our original image, I mean, they're in the same pose and everything, just a different angle on it. Here's another example from Henry Derez here where they did it with a cartoon scene. and they took this input image, asked it to change the perspective looking down, and it gave it this image. And you can see it's the same room, same color, same person, just a different perspective. Want to make a newer image look older? Give it a prompt like this. Change this photo to an image that looks like it was taken in the 1940s. Change the man's outfit to look similar to what a man from that era would wear. A few seconds later, we get a black and white image with some actual like old grainy effects on it and me wearing a completely different outfit. more representative of the era. And here's what it did for the 50s, the 60s, the 70s, the 80s, and the 90s. Probably pretty accurate, although I think they'd be in color for a few of these eras. Don't know why it made them black and white, but I think it got the clothing bright for the era. Want to put yourself on a magazine cover? Use a prompt like, "Put this person on the cover of People magazine sexiest man alive issue." And the nano banana obliges. Apparently, if I add to the prompt, make him posing for the camera in a sexy way. This is my sexy pose. Also, ignore the fact that it actually says sexist man alive. You can use this same technique to make movie posters. I love that I look kind of like Iron Man, but only from the waist up. And to make wanted posters. I didn't tell it to say this, but it says alias the grumpy gurup. And I'm wanted for crimes against sunshine and positive vibes. Reward a coffee and some kind words. I didn't tell the model to do that. You can also change the style of an image like you do with the OpenAI image model. Just give it a prompt like change the style of this image to look like a scene from Studio Giblly movie and it'll make an image that looks somewhat similar to a Studio Gibli style image. But what works even better is when you transfer the style from another image. So if I give it this same original image, give it a screenshot from a Studio Giblly movie, and then we can give it the prompt here that Google suggests to use, paste it in, transform the provided photograph of the two people into the artistic style of Studio Gibli, preserve the original composition, but render it with the style of image, too. And we get something that looks much closer to what you'd expect from a Studio Giblly style shot. The same technique also works if you give it a GTA 5 image and tell it to style it that way. Another really cool use case that I came across from Mad Pencil here was styling just a portion of the image. You could see she gave it this original image here, asked it to just make the ramen bowl and ramen look like 2D whimsical handdrawn anime. And this was the final result. Everything else stayed the same, but just the bowl of ramen turned into a cartoon. You could use it to add text to an image. Add the text subscribe to Matt Wolf in the background. The text should be behind the person in the photo. Here's what that created. Pretty basic. I asked it to make the text in the image much larger and make it appear like the person's in front of the text. And well, it just gave me back the same image. And I've tried this a few times now off camera and kept getting the same thing. So, I guess manipulating text like that is still sort of a blind spot for this model. probably something I would still want to do in like a Photoshop or Canva. Now, I've seen other examples where people were able to change the text on billboards. I was just having an issue a second ago with text, but let's see what happens when I try this. Change the text on the billboard that says billboard mockup to text that says check out futuretools.io. And would you look at that? It actually nailed it the first time. Let's take a closer look here. And everything's spelled right. It left the rest of the text the same as it was. So, sometimes working with text works really well. And for whatever reason, sometimes it's a little wonky. You know what the hardest part about keeping up with AI is? It moves way too fast. Just blink for a second and there's a brand new model. Like a crazy tool launch or some company making headlines for all the wrong reasons. Most people simply don't have the time to track it all. And that's why for this video, I partnered with myself the well, the Future Tools newsletter. It's completely free, goes out twice a week, and it's basically your shortcut to staying ahead of the AI curve. Every Wednesday and Friday, I round up the biggest AI news, the most useful new tools, and the stories that actually matter without you needing to doom scroll X or dig through a 100 blogs. And here's the best part. When you sign up today, you instantly get free access to the AI income database. It's a giant resource that shows you dozens of different ways people are making money with AI, complete with the methods, the process, and the exact tools you need to try it yourself. Totally free just for subscribing. So, if you're into AI tech or just don't want to fall behind, head on over to futuretools.io, drop in your email, and you'll be plugged into the smartest corner of the AI world, thank you so much to me for sponsoring this video. Now, let's get back to it. Another thing it's really good at is character consistency. So, if you're trying to create a whole bunch of shots for like a AI generated film or something like that, but you want the character to look the same in every single shot, this works beautifully for that. So, I went over to Leonardo, created this little green alien character. I think I like this one here. So, let's use this character. Now, if we pull in our character, this is our reference image. And if we use this every time, we should get the same character every time. Make the character in this image riding a bike. We've got our little green dude riding a bike. Here's our character shooting a basketball, our character holding hands with his girlfriend, him dressed up in a nice suit for a fancy dinner, working in a boring desk job that he hates, and finally getting a selfie with his biggest fan. But as you can see, we've got character consistency in all of these images. There's no doubt that that's the same character every single time. I also learned from Jared Lou here that you can actually annotate images. Give it an image and actually put the prompt that you want to happen directly inside the image. In this example, he used Canva, gave it a background image of like a living room, gave it a picture of a woman, said woman standing here, white rug on the ground, man sitting on sofa, plant in the corner, and then literally gave it the prompt, complete the prompts in the image. That was the whole prompt. And here was the output image. Now, there is a limit to how many images you can upload at one time into Nano Banana. But if you want to get a ton of images all into a single prompt, there's a workaround. You can actually go and use something like Canva, toss all the images into a single image and then prompt with that. So I quickly made this sort of collage of a bunch of different items inside of Canva. Now if I upload this image into Google's AI studio and say create an image of a desk that has all of these items neatly organized on top of it, look at what it generated for us. We got our drone, our speaker, our old radio, this glass thing in the background, some giant SD cards, but I mean look at this image. Look at the image I made in Canva here. It got every single element into this image, including this random old school little glass thing. That to me is really impressive. You can apparently do stick figure annotation. Give it two different images of two different characters and then draw the stick figures of what should be happening in that image. So you can see the person on the right is like jump kicking the person on the left. It should be these two characters. And it generated that exactly. Want to add branding to your products? Upload your logo and say something like, "Generate an image of a t-shirt with this logo on it." And we get a nice t-shirt that has that logo on it. But that's not really that impressive. Let's try create a perfume bottle with this logo on it. The perfume is called essence of future. All right, now we're getting somewhere. This is looking a little more impressive. Now I want to have like a model holding out this bottle. So I'll copy the new image here. We'll paste it back into our prompt box. Create an image of an Instagram influencer model holding out a bottle of this perfume to the camera. And we got an Instagram influencer holding out essence of future to the camera. And here's my perfume bottle sitting on the beach. If you want to change an existing logo that's already in an image, Punam here shows that you can do that as well. Here's a picture of Sundar with Google behind them, but then they swapped out the image with OpenAI, Microsoft, and even XAI, but the foreground is all the same image. Need a business card made up? Mock up a black business card with this logo on it. Uploaded my Future Tools logo. It should have my name on it, Matt Wolf, as well as our URL. Add additional stylized elements to make the business card look professional. And well, here's the new business card. It actually looks pretty clean. I didn't really give it a whole lot of other details, but it add all these like style elements to it to make it look nice. Need a website mockup? Give it a prompt like mockup a landing page for a site called Future Tools. Use the logo in the image provided. The site should be tech themed and focused on promoting the latest AI technology. And I gave it my logo. And here's the landing page mockup it gave me. Honestly, not too horrible. It's not amazing, but it could give you a good idea for like a color scheme to go with. I don't really like how this sidebar is so tight to the image down here. But it did a pretty decent job of making something that looks interesting and quite honestly not too far off from what my site already looks like. Just go to futuretools.io io. O, if you need proof, need to create an online banner ad to promote your product, like for instance, your latest fragrance. Create a banner ad that promotes the fragrance Essence of Future. Be sure to use the provided image within the banner ad. Here's the banner ad it gave us. Essence of Future. Unlock tomorrow's scent. Shop now. It's a good ad. It could do YouTube thumbnails as well, although I personally haven't been too impressed with the thumbnails it created. It can do it for you and get a pretty close starting point, I guess. If I give it the prompt, create a clickbay YouTube thumbnail that includes the person in the image provided, they should be giving a surprised face. The thumbnail should be for a video about how to use AI to make wild images. And here's the thumbnail it created. And it's honestly not horrible. It might actually work. Definitely not up to my quality standards, but again, actually a little bit better than I thought it was going to do. Nano Banana can help you with landscape design. This image that you're seeing on my screen is an actual image of my backyard. We actually just had a bunch of construction done to build this giant wall. And in the process, all of our landscaping was destroyed to bring in all the heavy machines. So, we have like this dirt back lot right now. I took this image into Nano Banana and literally just put beautiful landscaping on this yard. And here's the image that it gave back. And I can go give this to like a designer now and say, "Hey, help me make my yard look like this if I want." This is an actual image of my pool area on my property. Again, destroyed by recent work that we had done. I asked it to redesign it and make it look prettier. And here's what it designed. And it looks pretty good, honestly. Apparently, these flowers grow straight out of my wall. But other than that, it's a really decent look to it. This can also be used to redesign what's going on inside of an interior room as well. Jacqueline Consleman here shows off an example where she gave it this image and actually circled an area on her wall and said, "Put wallpaper on this part of the wall and then got a handful of variations of what different wallpapers on just that wall would look like." My buddy Balav showed off this really cool use case. You can upload screenshots of the real world and ask it to annotate stuff for you. Here's the actual prompt. You are a location-based AR experience generator. highlight and then you know include a point of interest in this image and annotate the relevant information about it. Here's an example of what it did with the Transame pyramid in San Francisco, the San Francisco Ferry Building, the Palace of Fine Arts, and another angle on the Palace of Fine Arts. I actually tried this one myself in a previous video with Petco Park, and it actually highlighted Petco Park downtown here. Simon here showed off that you can give it a prompt like, "What does the red arrow see?" give it an image like this of the San Francisco Bay and just point a red arrow and it gives you an angle of what it would see from the red arrow. Now, I'm sure it's not 100% accurate. I I don't know if this is exactly what you would see in the background, but pretty dang close and pretty dang impressive. This was another one I had to test myself. So, I did it with Petco Park again. Give me an image of the ground view from the perspective of the red star. I gave it this image of Petco Park with the red star right here and it gave me back a perspective of what it could look like from that position. Now, I know from experience and knowing downtown San Diego very well that this isn't quite right, but it is pretty close. A lot of people have been doing these really cool isometric images of buildings as well. Harath here gave it this image and gave it the prompt make image daytime and isometric temple only. and it gave them back this isometric image. Like super impressive. Another example from Ilia here. Make image daytime and isometric. This was the image they gave. Here's the isometric they got back. I asked it to make an isometric image of Petco Park. Exclude the surrounding buildings. Gave it an image of Petco Park and got this image back. And it's I think pretty spoton. Here's the real image of Petco Park. Here's the isometric image. Swapping back and forth. I think it pretty much nailed it. If you're into coloring, you can make coloring pages directly from within Nano Banana. Here's an image that I made in Leonardo. Let's toss this in. Give it the prompt, convert this image into a black and white outline for a coloring book. And now I get an image of myself flying on a dragon that I can go and color in. You can also take like a child's drawing and tell it to make it more realistic. I'll just give it the prompt, convert this into an image that looks real. And it kind of made it look sort of like a realworld like painting or like string or something. By tweaking my prompt a little bit and using create a real world photo of a house based on the image uploaded, I get something a little bit closer to what I imagined it would do. Now it still kind of looks like chalk drawing on the sky, but the house actually got more realistic, which I think is kind of a cool aesthetic. I saw from Henry Dere here that it can actually do a pretty decent looking behindthescenes image. So, I'm uploading a shot from one of the Marvel movies and saying, "Create the behindthe-scenes view of them shooting this movie." And here was the original image that I uploaded. And here's the behind-the-scenes shot that it gave us. That is better than I was expecting. Again, the original. We can see they're all in these poses here. And when we jump over, they're all kind of in the same poses. I think Jeremy Rener got deleted. You got the lighting, the green screens, the cameras. Looks good to me. I've also heard that it's very good at making sort of deconstructed views of an image. So, let's try this. Create a deconstructed version of the Apple Vision Pro from this image. And here's what it created. I mean, I have a feeling that that's not really what the inside looks like at all, but the AI model doesn't know what the inside looks like, so it did its best guess. And according to Lewis here, nano banana is really good at creating nano bananas. See, here's an example of a nano banana. Here's some more examples of nano bananas and the word strawberry spelled out using nano bananas. All right. Now, I've given you a pretty exhaustive list of all sorts of things you can do with nano banana and it's probably just scratching the surface. So, if you have ways of using it that I didn't talk about, make sure you leave them in the comments so others could learn from the value that you have to offer. But I'm not done yet because there's some really cool stuff you can do with the nano banana images that you make by pulling them into other tools. So, I want to share some of that really quickly. Remember how earlier we created an isometric view of Petco Park and we saw some examples of other people creating isometric views of places? Well, you could take those images and upload them to a tool that will convert it to a 3D object. Microsoft has what they call Copilot 3D and you can upload the image here. You definitely lose some of the details here, but this is a 3D image of Petco Park sort of. However, the 3D generator that I prefer is actually Meshy.ai. I've always been pretty impressed with what they can generate. Now, Petco Park is well, pretty complex, but we're definitely getting a better 3D model of it than what we were getting out of Microsoft's version. But, you know, still not great. There's still a ways to go with getting finer details on some of these 3D objects. But, a fun thing that you can do with some of those isometric images you make. If it's a fairly basic isometric image, you can pull it into something like Meshy, convert it into a 3D object and then potentially even 3D print it. But again, I think Petco might have just been too complicated of an example. Remember how we made our perfume bottle earlier with the influencer holding it out? Maybe I want to turn that into a video that I use on Instagram or something like that. Well, I can head over to Cling AI, give it the bottle as our first image and the influencer holding it up as our second image, and we get a video that looks like this. I mean, pretty professional looking. I like how you can even see movement of the liquid in the bottle as she's holding it up and moving it around. Remember our deconstructed Apple Vision Pro here? Well, if I jump into Cling, give it our constructed version and our deconstructed version. Here's the end result where we have our version sort of deconstructing itself. I also took our alien friend here, brought him over to Clling and generated a video of him riding a bike. And then I took the image of him playing basketball and animated that. I mean, Clling kind of messed up the movement of the ball a little bit, but you can see how you can get animations with consistent characters that look like the same character in every scene. But what about the intro? How did I do that? I recorded a video of myself saying the intro to the video you're watching right now. And then I took a screen grab of me just presenting that intro. This is just a screen grab from the video. And I said, "Change the background of this image to a lush green rainforest. Leave the person in the foreground the same." And you can see it's the same exact image with the green background. I then took the same screen grab and said, "Change this person in the foreground to look like a Rick and Morty character. Leave the background the same." This was the output, the second character you saw in that intro. And then I said, "Change the person in the foreground to look like a monkey human hybrid. Leave the background the same. And here's the monkey you saw." And then I jumped over to RunwayML here. And with RunwayML, you can actually use their generate video module and then make sure this is set on act two. And in act two, you give it a driving performance. You give it a video. So I just fed it my whole intro video of me giving the intro. And then you just give it an image and it will animate the image saying the things that you said in your driving video. So you can see I got a whole bunch of outputs. I got a video of me giving my intro in front of a green background. People are doing some absolutely wild things with the a version of me as a Rick and Morty character. People are doing some absolutely wild things with and a version of me as a monkey. People are doing some absolutely wild things with the I just sort of spliced those various videos together and that's how you got the intro. Nano Banana plus runway act two. I am absolutely having a blast with Nano Banana. This video was not sponsored by Google whatsoever. I have just been using this tool probably more than any other tool recently. I've been kind of obsessed with using it to make edits to images and then turn them into like videos and things like that. It has been so much fun to use. And I just wanted to make this breakdown and make a pretty exhaustive video of all of the cool ways I've seen it being used so far. And again, it's just scratching the surface. There's so much more you can do with this. And if you've got ideas that I didn't even share, make sure you either comment below this video and let everybody else know or tag me over on Twitter or threads or Instagram or Facebook. I'm kind of on them all now. Tag me with what you created and what was different and let's share the word about the various cool things we can do with Nano Banana. Cuz again, this is so much fun. But that's what I got for you today. I hope you enjoyed it. I hope you learned some new techniques. This is amazing to me that you can use this completely for free right now. Like that blows my mind. It's out there. It's amazing. It's fun. It's free. Go play with it. Thank you so much for tuning in and nerding out with me. This has been a really fun nerd out session. And hopefully I'll see you in the next one. Bye-bye. Thank you so much for nerding out with me today. If you like videos like this, make sure to give it a thumbs up and subscribe to this channel. I'll make sure more videos like this show up in your YouTube feed. And if you haven't already, check out futurtools.io io where I share all the coolest AI tools and all the latest AI news. And there's an awesome free newsletter. Thanks again. Really appreciate you.