Transcription
Welcome to another Google Gemini AI video tutorial.
There are many websites that now offer nano banana through API, but if you have a Google account, you can access the official Google platforms by searching for Google Gemini. You can then open the gemini.google.com website. Log in and in the top left, you will see the model 2.5 flash, which is the name of the nano banana model. Then you can prompt for an image or you can upload an image and edit that image. So let's upload an image like this portrait of a woman. And then you specify what changes you want like changing the color of the t-shirt to red. After that you submit using this button and it will take a few seconds. I saw it generate images faster than chat GPT and you get a new image with the t-shirt red just like I asked.
Google uses something called synth ID from Google DeepMind to embed invisible digital watermarks into content generated or edited by its AI tools. The watermark is imperceptible to regular eyes and also includes a visible marker like the small star you see in the corner. Google uses this symbol consistently so people can recognize AI generated content at a glance. The invisible part of synth ID is the real verification watermark while the star is just a visual cue for humans. You cannot remove the invisible watermark by cropping or filtering. You can crop out the visible star but that does not remove the hidden watermark. You can use this download button to download the image.
If I compare the original image with the edited image, you can see it did a good job editing. It just adds that annoying star. Not sure why they add it if they already have the invisible watermark.
The other way to access it is through Google Studio. You can open the site aistudio.google.com and you can collapse the sidebar to see what it offers. You can use this chat and on the right you can select different models. As you can see, we have the nano banana here. So you can select it. Then you can add images from here. And I will add the same image so we can compare. Click acknowledge if it asks you that. Then add the prompt and press the run button. It works similar to Gemini just with a different interface and more settings. Useful for developers. You can use it as a temporary chat or save the conversation to Google Drive. You can also enable that function if you want it saved. It also has a download button. We got a similar result from both. The only difference is that on Google Studio, the star is a little smaller.
If you find that star annoying, you can use tools like Photoshop to remove it. You can use a selection tool to make a selection around the star. And in the new version of Photoshop, there is a remove button. Older versions of Photoshop also have remove tools or clone tools. In just a few seconds, you can remove the star. You can also add your logo on top of the star to hide it.
Let's continue now with 50 use cases for this nano banana model. Use normal language and descriptive prompts for better results. I will include all the prompts and images generated in this video on Discord so you can download them for free. Check the video description.
You can remove unwanted subjects from a photo. For example, maybe I don't want this kid in my photo. I just explain that in my own words like remove the kid running behind me and fill the background naturally. Then I get an image without that kid. So, it is quite good at removing objects and people from images.
It is also quite good at removing or changing the background. In this case, I wanted a white background and also to add some shadows. It worked great for that. I can come up with more changes like making the background black and I get a new image with that change. This one still has that watermark star, but because it is white on white, it is not visible.
You can place a subject in a new location like I did with this man who is now in Paris in front of the Eiffel Tower at sunset. The face is quite consistent and very similar to the original image.
I also like to add a new object or element into a scene like I did here when I added a cake. Because I didn't mention what type of cake I wanted, I got a random color and style of cake. So add more details in your prompt if you want more control. For example, I asked for a bigger pink cake and I got the cake bigger and pink.
Then I wanted to add a clown, but it refused to do it. I tried to go around that, but it still didn't want to. So maybe I need to prompt in a different way. I asked why it didn't want to, and it seems it associates clowns with negative emotions. Then I added an old lady behind her, and it did that for me. After that, I asked for a clown face, and again, it refused. If that happens to you, just take the last image and try a new chat. Sometimes it gets stuck in that chat and refuses to do it. But if I gave the image in a new chat, it did it without a problem.
You can change clothing or an outfit in an image. For example, I wanted a red dress for this woman and it changed that for me quite well. Then I wanted a pair of round glasses to match the outfit and the result looks quite nice. You can also change the hair color and style. It is better to do one change at a time like I did here so it does not get confused with too many changes. The instructions can still be quite detailed and long.
For all the ladies out there, you can apply subtle makeup or retouching or visualize ideas before you actually do the makeup so you know what works best for you. I believe in the future we will have all kinds of mirrors with AI that will do that live.
Wouldn't it be nice to change the weather to fit our mood? Well, now you can. At least in the AI world, I have this bunny illustration and I wanted to make it a rainy day and make the bunny sad, so I gave a detailed prompt. I got this image, which is pretty nice, but the sunlight was still hard to remove. Even though I prompted for no sunlight, it still added some. Redoing an illustration like this manually would take forever, so AI is quite useful.
You can change the season of a landscape, like I transformed this summer tree into a snowy winter scene. Then you can continue the conversation and make it autumn as well. I like to do that sometimes for the first and last frame of an AI video. So I can then create a video transformation from one scene to another. That way they are similar and the transition is smoother.
Here I tried to make this person look 20 years older, adding some gray hair and wrinkles while keeping their identity. The result looked a little older, so I asked for older, like 100 years old, and it gave me this version. I am not sure if that is how it would look at 100 years, but who knows? Then I wanted to try 1,000 years old, and I got more of a mummy or vampire look, more in a fantasy style. I then asked for it to be more realistic, not an illustration, and it responded that maintaining realism at such an extreme age challenges its capabilities, so that is why I got that kind of image.
Now, you can upgrade your food photos to look more professional, like I did with this burger. it tends to add a plate under it. Even for burgers, when I asked for Michelin star style, so maybe you can describe better how you want the food to look.
You can convert your photo to different cartoon styles. I wanted this 3D cartoon, but I wanted it to look more similar, so I asked a few times and got it closer and closer to my image. Just chat with it and ask for more changes until you like the result. You can also convert to a painting style like I did here. You can try different styles like watercolor painting or impasto painting and then when you like the result you can add a vintage frame. It is quite fun to play with it. Now I have two watermarks like inception with a watermark in a watermark in a watermark.
If you have old black and white photos, you can use Gemini to colorize them like I did with this photo. You can go further and change the outfit or improve more if you want. Or you can restore old photos, remove the scratches and fold marks, and sharpen the faded faces. It did a really good job at fixing those.
You can also use it to remove watermarks, cables, and all kinds of distracting stuff. It is pretty good at converting sketches to a final design, though with logos, it wasn't as exact or as creative as I hoped. Still, it did an okay job. Then I asked for a different background and a more futuristic look, and I kept asking for more versions. The versions got more and more detailed until I ended up with a game-like logo. At some point, it lost track and made a totally different logo. So, I am not sure if it forgets the context after a few tries or what is happening, but you can also take the logo you like and make changes on that version.
You can create poster or flyer images. I asked for a ratio there, but Gemini is not so good with ratios and sometimes fails to get the exact one I want. In this case, it said that the 3:4 ratio was closer to what I asked for.
You can mock up a billboard or sign and place your advertising in different environments, and it kept my image pretty consistent. You can make it even more complex, like I did here with the billboard seen from inside a car. The details are kind of lost if the image is really small. For example, it messed up the word innovation a little since it was very small, but it still seems to do better than chat GPT at least.
I like to place my logo on different products and Gemini or Nano Banana, however you want to call it, is quite good at it. Here is an example on a black mug. Then I tried a more complex example with an energy drink can adding extra text and ice cubes and also on a sign. It depends on the logo. If it has really small text, it might mess it up a little, but usually a logo should be clear and simple anyway. And for that, it seems to work fine.
Not only can you place the logo on a generated product, but you can also place it on a specific product. For example, I wanted to place it on this exact photo of the man with a black t-shirt, and it did a pretty good job. Then I tried something more complex. I wanted the logo on a specific coffee bag and also on an exotic leaf. And I asked it to integrate everything together. It had no problem handling all those images. After that, I added a hand reaching to grab the bag. This is quite useful if you do first frame and end frame AI videos.
Let's say you have a design that you want to visualize on a product. I added that design and the paper bag I wanted it placed on and I got this image but for some reason it made it portrait instead of square. Probably because it saw the design was in portrait mode. I asked to make it square ratio but it failed. So it is kind of stupid when it comes to ratios. I asked again hoping it would fix it and it failed again. Maybe I should try a new chat for that but it is probably faster to do it in Photoshop. Then I tried with another design and it placed that too but it didn't fill the entire bag area. I asked again to fit the entire front area and not leave empty spaces. And then I got better results that fit. I gave it a different design this time but now the design was square not portrait ratio. And look at that the design is square ratio. So try to upload the images in the ratio you want in order to generate the right ratio. I wanted to extend the design to fill the entire bag and the result is quite good.
You can try with all kinds of products. For example, I have this armchair in this pattern and I wanted to cover the entire chair with that texture. It did a better job than I would do in Photoshop with the warp tool which would take a while to make it fit and get the right lighting.
So now I want to make a YouTube thumbnail. I added a 16-9 ratio image for the background, hoping it would trigger the right ratio. Then I used this photo of a surprised man and this octopus image. I asked for a YouTube thumbnail and gave details on how they should be placed and also to add the text survive. The result is quite nice, though I could still make some changes. I wanted to change the lighting to golden hour and adjust the lighting on the man and the octopus as well. I also wanted to move the text on top, but instead of moving it, it duplicated it. Then I asked to remove the bottom one, and now it looks more like how I envisioned the thumbnail. It is probably easier to remove the text and add it exactly where I want with the font I want in Photoshop. But for the rest of the image manipulation, it is definitely easier than Photoshop because it combines the right lighting and shadows, which are very timeconuming to do in Photoshop.
You can also create memes or edit them. For example, I replaced the woman in red with a witch in the famous meme, and you can come up with all kinds of ideas. Then I asked to add some text to the bottom. If you don't tell it where to put the text, it will place it in a random position.
You can generate all kinds of visual aids, images for pneumonics, flashcards, and educational material to make it easier to remember, like I did here for raining cats and dogs.
It is also useful for interior design visualization. I have this wooden door and maybe I want to replace it with a new door, but I want to see how it will look before I buy it. This way you can get a quick idea of how it would look. Then maybe you want to paint the walls in a new color. I am not sure how accurate the colors can be, but still you can get an idea of how it will look. You can add furniture, change things, and have some fun.
I have this modern home here and I wanted to create an architectural 3D isometric view of that house. The result is better than I expected. It picked up a lot of the details from the image and I am quite impressed.
What about fashion design and trying on new clothes? Well, it seems it can do that too. I put a yellow dress on the woman, but it had a small mistake on the sleeves which were longer than in the original image. I then switched to medium portrait and asked to reduce the length of the sleeves, but it didn't work. In my mind, I yelled at Gemini and it apologized and fixed the sleeves, making them more similar to the original. Then I asked to add a purse. It did that, but when I asked to place it in the other hand, it didn't work. I asked again nicely, hoping for a fix, but the purse was still in the same hand. So I thought maybe I should ask what is wrong like I am a psychologist for an AI. What is wrong my dear AI? But in my mind I wanted to smash it. It seems it has trouble knowing left and right from the perspective of the image. So instead I should say the side of the image instead of her hand. I asked for the left side of the image and Eureka. It finally did it. So you just have to know how to ask. Try not to get mad at it. When the robots come, you will be safe if you don't upset them. Or maybe not.
I have this woman holding a perfume bottle. Wouldn't it be nice to be able to place my own perfume bottle in her hand? Well, if you ask to remove that item and add your item, it seems to work. I got this image and it is not bad at all. Here is another example with a hand holding an apple and a white mouse. I wanted to remove the apple and add the small mouse in the hand and it worked quite nicely. Then I replaced the background to change the story to some secret project. But the hand looked too simple. So I asked to add a white glove. Look at that. It added the glove and still kept the mouse. Better than I expected. Then I asked to make the mouse jump toward the viewer. And I think I can work with that. I can continue it with an AI video.
From here you can generate a character sheet image for your characters. useful for those who do 3D character design and want to see them from different angles. I got a front, side, and back view that is pretty similar to my original character. Then I asked to change the outfit to a Halloween wizard outfit, and again, it did a good job. It is not always perfect and sometimes mixes up right with left, but it still saves a lot of time and is quite useful.
If you want to generate an image in a certain style, it is easier to use an image in that style instead of writing long prompts that the AI might not fully understand. For example, I wanted the same art style and color scheme, but with a castle instead. Of course, you can change the color theme, but the style is the most important part here. Here is another example with something more 3D. And again, the style was captured in the image generation. Another example is when I wanted a wolf with axes behind it. And look at this beautiful shield I got. Then you can make small adjustments like colors and details. This is my favorite way to prompt in certain styles.
You can combine multiple images to create a realistic family portrait of a man and a woman as a couple holding the cute bunny together. I added more details about the style because the styles were different. So instead of getting a random style, I had more control. The result is quite nice and the bunny looks similar to my original image.
Another function I use a lot is to convert line art to 3D render or sketches to 3D or even realistic images. Look how well it created the 3D render. And if we compare before and after, it kept the character pretty consistent. Even if you don't want to use AI for generation for your own design, you can still use it for inspiration, like getting color variations quickly before spending hours on rendering. You can also create coloring pages for kids and adults from existing images or create line art for engraving in other projects. Look how well it did the line art for this bunny character. I also tried with a woman to see if it can do it and the result is pretty nice as well. It is worth a try considering that you can generate a few free images every day and many more with the premium account.
Another fun way to use Nano Banana is to create caricatures of yourself or your friends. Great for gifts or greeting cards. I got this caricature of that woman and it looks quite nice.
You can also make personalized greeting card images. For example, I took this woman and asked it to place her in a realistic winter snowing scene with falling snow and a festive atmosphere. I asked to dress her in a classic Santa outfit and to add the text Merry Christmas on top in a bold festive font. I was asking for a lot of changes, but if it cannot do them all in one go, you can adjust step by step. Or you can simply generate custom greeting cards without an image. Just tell it what to include, what text to add, and what design you want, like I did here with this vintage card.
Another cool thing it can do is replace text. Like you see here, the text is 3D and at an angle, and it was able to recreate it in the same style, which is quite impressive. You can also try it with different languages.
Besides replacing objects and backgrounds, you can also replace a character. In this case, I wanted a panda eating bamboo leaves instead of the bunny, and it kept the same cute style and 3D render so it understands what is in the image when it does that replacement. You can also remove the panda entirely if you want and keep the background for different projects, or make a video with the first frame empty and then the end frame with the cute panda sitting in the middle of the scene.
Using the same bunny image, I asked for a cute sticker style image on a black background and also added a hat and sunglasses. It gave me a pretty nice sticker. So, you can use this method to create an entire sticker sheet by making different poses for your character. You can also make emojis if you need them for Discord or other projects.
You can also make a photo look vintage. For example, you could use the same woman but add a vintage effect, but I also wanted it to fit better with that period. So, I asked for a different hairstyle and a different shirt to match the time.
Besides simple makeup, you can also try more complex looks like Halloween makeup or different holiday themed outfits from Christmas to Street Patrick's Day or any other holiday. Speaking of Halloween, you can transform your house photo into a haunted house Halloween setting. It might give you some ideas on how to decorate for Halloween, or you can simply share it on social media.
Another fun way to use AI is to create superhero images of yourself or your friends. I got this interesting image with just a simple prompt. It looks really cool, don't you think?
For those who are into user interfaces, you can create game interfaces or simple interfaces for mobile or web. Just give instructions on what you want and you can get a lot of interesting ideas for your next project. And of course, you can create game elements like icons and buttons like I did here with this set of four icons. You can do more, but the more you add, the smaller the icons become, and the less detail they have. Four seems like a nice number for a set. And you can also specify in the prompt what each icon is and where it should be placed, like top left, bottom right, and so on.
I tried to see if it can do Lego brick sculptures and it works for many images but not for all. For this bunny, it worked great and I got a lot of details like a complex Lego masterpiece. I also tried with this woman and it kind of adapted her to make it work. You can try different prompts and maybe get better results. Then I tried with this house and for some reason it didn't want to do it. It only added some pieces on the floor and in the brown trees.
I also like to make 3D renders from simple shapes, logos, or text. For example, with this Pixarroma logo, if you don't know about Pixarroma, make sure to check the Pixarro YouTube channel. I created a 3D golden logo from that simple shape. It can be quite useful for creating all kinds of designs for branding and advertising. Then I tried with some simple text. Try different prompts. This could have worked better if the text had a different color, maybe made from a gemstone. Then with the same text, I tried something that would fit better, made from gold with snow and a Christmas look, and it worked great. No control net was needed, just prompting for what we want. That seems to be the future.
Here, I tried to make a stained glass design from this gnome image, and the result was quite good. It can be useful for different illustrations or crafts. I also tried it with this illustration, but the result was a little too flat for my taste. I am sure it can do better.
Let's try something more surreal like creating hybrid animals. Here I mixed a lion and a butterfly. It is not perfect but it makes a nice illustration. Then I tried a long prompt for something more complex with a lot of details and the result was this unique creature that you have never seen before because it was just invented by mixing different animal parts. It actually put them together quite nicely. Here is also a combination of a zebra and a rhino. So try to push the limits of this nano banana model.
Here is the last use case, but there are many more things you can do. These are just the ones I use more often. Architectural visualization is quite popular. I had this concept design of a building and I converted it to a 3D architectural rendering while keeping the proportions and design. The consistency is quite impressive. Then I asked to show the same image at sunset since golden hour always looks nice. After that, I changed the camera angle to show the top of the building. You can also add more elements like I added this helicopter in the image. And since we are at the end, maybe add the text the end over the image.
I want to thank AI Titans for the support. Thanks to Sebastian Anthony, Uptown Funk, and Thomas Brown for your support, and thanks to everyone who subscribed to the membership. I worked a few days on this video, so if you found something useful, leave a like and a comment to help with the YouTube algorithm. Don't forget to check Discord for prompts and resources. Have a great day.