Transcription
Google's AI Studio might be the best kept secret in the AI world. Not only can you chat with AI for free with really high limits, you can also build full AI apps using Google's AI Studio, and it's 100% for free. And if you want to take what you build, bring it into something like Cursor, and continue building out something more complex, you can do that, too. So, it's a great starting point.
So you'll see in this video everything that you need to know about how to use Google's AI Studio. So let's start here in the dashboard and I'm going to talk about some of the different functionalities that AI Studio has and then we're going to dive into each of those individually. So you're going to see how to use it.
So AI Studio is really something Google has built for developers. Although even if you're not a developer, it's still really easy to use and take advantage of all of the free tools that they have here. but it's meant for developers to prototype and build prompts and build apps and test out the API and then when they're ready to then integrate it into apps outside of AI Studio.
To get to AI Studio, you can just type in AI.dev and then it'll take you to this AI studio page. On the left side, we can see the different tabs that we have available to us. We have just a chat tab, we have a stream tab, we have a generate media, and a build tab. Those are the four main functionalities, but each of them has a lot of detail to get into.
So, we're going to start with the chat tab. So, here on the chat tab, there's a few things to be aware of. Here on the right side, you can choose all of the settings for our chat. So, you could think of this kind of chat GPT or even Gemini. Gemini is Google's chatbot, but this gives you a lot more control.
So, here under featured models, for example, we can see they have the nano banana is what they're calling it. It's Gemini 2.5 flash image preview. And you can see there is pricing for what it would cost to use the API. But here inside Google's AI studio, it's free. So you don't have to worry about that. But this is the pricing if you were to use it inside of an app outside of the AI studio.
Then there's 2.5 Pro. This you could think of this as a comparison to GPT5 in terms of power. Gemini 2.5 Flash. This is still really good, still a reasoning model, but it is a lot cheaper and faster than 2.5 Pro. So, a little bit less smart. And then 2.5 flashlight goes down even more. And you can see the price differences. If you're using this in an app, it's 10 cents for input, 40 cents for output versus pro is if you're using it depends on how many tokens you're using, but $1.25. So, it's more than 10 cents less for or 10 times less for flashlight than 2.5 Pro for input. And then for output, even more than 10 times less. So, way cheaper to use this model, but again, you're not going to get quite the results.
So, you have model selection. You can set system instructions. So these are all of the things that if you were building an app and wanted to integrate an AI, you can set all of these things in the API. So let's say system instructions, we want to say everything in lowercase. Just something really simple, but you can do that. And then that will be there as system instructions. And then let's say we want to do 2.5 flash model. And then everything here can be customized as well. So thinking mode on or off, how much thinking budget you want to have, which is tokens, whether you want to add Google search groundings. So basically, it'll get sources from Google search or not, whether you want URL context, and then here, if you click get code, you can get the code to use the settings inside of an app.
Okay, so now that we've set this up, we could send out a prompt. So what if we just do something simple, tell me a short story, and we're going to run it. And one thing that's going to be cool is because we have system instructions as write everything in lowercase. Everything here is coming in lowerase. So that's how you can influence system instructions. Temperature this is creativity. This is one right now. You can make it more creative or less creative depending on what you want. The default is one. But yeah, that is the basic functionality. We have a chat. We can respond. Tell me more. And you can see how quickly this comes because we're using the flash model. If we were using the pro model, it would definitely take longer to get these responses.
So, there's a few cool things to be aware of. So, one, all of these messages, we can edit them, and then we can stop editing, and then we can basically edit the messages as we're going to keep testing. We could go back to this message, tell me more, and I could edit it and say something completely different. So, I could say, write one more sentence. And then what I can do is either stop editing and save that or I can rerun it from that turn. By doing that, we're going to overwrite the message that we just got. So now we're getting another message. It's just rewriting it. And you can do that on any of the messages. You could also just say, okay, we're going to keep it, but I want to rerun it again. So this is a really cool way to be able to do that.
We have a few other features. You can delete messages. You can branch from here. So, if I click branch, what that's doing is that's putting us into a another chat. And then you have you can click here to go back to our original chat. So, basically, you can branch off and see the conversation in in multiple ways.
A couple other things you can do now that we've started this, we can save this prompt and then in the future, you can go to your history. We can click on that and then everything is going to be there with the same system instructions. So you can continue working on a prompt kind of conversation that you were working on before. You can also share the prompt with other people. This is similar to sharing like a Google doc.
We can go into compare mode and we can change the model on Flash versus Pro. And we could say continue. And this is a really great way to compare the different model responses. So this is especially if you're building an app, you want to think about what is the model that you want to go with. This is a really great way to do that. So the compare mode is cool. You can do two models at once and see them stream in at the same time. And then I can exit out of that. And then there's just new chat as well, which will do a new chat with the same run settings that we have here.
So that's an overview of the chat mode here in AI Studio. Again, it's completely for free. If you just want to use this instead of using chatbt or Gemini, you can do that. And it would take a lot to get to the point where you are hitting their limits.
Now, one thing that I've used this before for is let's say I am having trouble in cursor or something and I'm building an app. There's tools that you can use to export all of the code into a markdown file from your a code repository. And because the context window is so high in these models, you can upload an entire codebase or almost an entire codebase and ask it to debug something, help you with that. And that's something that I've used AI studio for before. So, that's a really great use case as well.
So now let's jump into the stream feature here. So what this is doing is it allows you to talk to Gemini. So this might be a little bit harder for me to test here while I'm recording the video, but I'll do my best and see what's possible. So again, you can choose the model that you want to run. So this is just a great way to test all of the models. You can change the voice language. Everything here again is going to be the settings that you can do in your model.
So let's say I start a prompt. So, let's just say hi and see what happens here. Hi there. How can I help you today? Cool. So, you can see I'm not going to be able to use the audio recording feature right now because I'm recording video, but you can just type in messages and then the model will respond. You can see how quickly it responds as well. I am doing well. Thank you for asking. How are you doing? Yeah, like pretty instant it's responding.
Let's say I wanted to say help me learn Spanish. Great. I can definitely help you get started with Spanish. First, do you have any prior experience with Spanish or is this your first time learning the language? It's my first time, so I'll just do one more message. Okay, perfect. We can start with the basics. A good place to begin is with greetings. For example, hola means hello. Buenos means good morning. Buenos means good afternoon. And buenos means good evening or good night. Would you like to practice these greetings? Okay, you can see how powerful this is. This is making me think about different apps that I could potentially build with this voice feature. It's actually even more impressive than I remember the last time I checked. I think maybe their new models have improved this even more, but that was cool.
So, this is something again you can do. you can get the code and these are all things that in a second we'll show you how you can start building your ideas inside of AI studio or if you're doing it with lovable or bolt or cursor you can tell it to use these APIs to integrate into your apps as well.
Next we have generate media. So this is where you can generate images, speech generation, music and video. So we'll go through these really quickly. So first we have the nano banana model. This is putting us back into chat because it is a chat model, but it generates images within the chat. So, we're back into chat. The only difference is now we're on that nano banana model. And I could say, "Design me a birthday card with a dog jumping out of a peak." Let's just see how well it does with this. So, it's going to give you a text response and it's going to give you an image response. So, you get the text response first and then here's our image response with the birthday card. And all of the other features I talked about before are all relevant here. Everything rerunning, editing, system instructions, all of that is possible with nano banana. You do have a few less options of what to customize with the nanobanana model. Different models will have different options there, but that's the nano banana model here. Generate image.
Then Google has another image model called image. So, this is just a pure image generation model. It's not like nano banana where it's a conversational and you get like the chat response as well. What's nice about this is you can change the aspect ratio just directly here. You can get more number of results. There's just more kind of image specific features here. You can even change the resolution. It is free as well, but you can see it has a free quota. So if you go beyond this limit, you're going to have to use the API. But as I mentioned, it is free to try all of these.
So, we could generate an image of a countryside with cows and horses. And we can choose our image aspect ratio. Let's do 16 by9. And let's do two results. And we can go ahead and run that. This is also going to take a little bit longer. But the results I've found are a little bit better than Nano Banana, but Nano Banana is actually that was really fast. But Nano Banana is a little bit better at editing the images cuz you can do it in a chat. So, yeah. So, we got two images super quickly and it was free to do that. So that's a cool feature that we have built in.
Go back to generate media speech generation. So here what I think is really cool about this is you can do multi- speakeraker audio. So we can have two speakers. We can create a script. Please read the following in a podcast interview style. We can have speaker one, speaker two. We can have speaker one settings, speaker two settings in terms of their voice. And then we can run that. And then we can get a voice audio with multi speakers which I think is really powerful. So let's test this out. We're seeing a noticeable shift in consumer preferences across several sectors. What seems to be driving this change? It appears to be a combination of factors including greater awareness of sustainability issues and growing demand for personalized experiences. So you can do this with a lot more dialogue. You basically just add keep adding messages. Speaker 1, speaker 2. And again, what's so cool about this is can all be done via the API. You could create your own app that generates podcasts with multiple speakers. for example. You also can do it with one speaker as well and just get the responses here. But I think the multie audio is, in my opinion, what makes this really cool. And then again, you can choose between the Pro and the Flash model. The Flash model is just going to be quicker. We can test it out really quickly. We're seeing a noticeable shift in consumer preferences across several sectors. What seems to be driving this change? It appears to be a combination. And I'll stop it there. You want to listen to it again. I thought that the quality was pretty similar because I edited out the loading time. you can't see, but it the flash was quicker to get the response.
Next, we have LIA real time. This I haven't played around with it as much. I think it's like an AI music generator. So, what I'm curious is if can we prompt the kind of music that we want and then it'll generate it over in the preview. That's what I'm most curious about. All right, so it updated and we can see all the sounds changed and now they're all sounds that are more like relaxing. And let's press play. [Music] Okay, so that did work. And so that's what is possible here in the music AI generation. You can prompted to ask for a specific type of music. You can also ask for I was looking in the documentation and you can ask for like specific like scale models, C major, D flat, that kind of thing. You can ask for different beats per minute. There's a lot of different like more musical types of things that you can ask for. But you could also ask for the genre, the instruments, the mood and description and put it here into the chat and then get an AI generated music player here.
And then finally, we have VO. This is where you can generate videos, which is really cool. It's getting like cooler and cooler as we go. Next, we're going to get to the build tab, which is more like a free version of Lovable and Bolt. So, we can generate videos with VM. This again is going to be rate limited, but you can do it for free. Generate one result here. You can choose your aspect ratio, the length, frame rates, output resolution, negative prompt, but let's just describe a video here. So, we'll just do something simple. A dog running along the beach. We will send that off. This is going to take a little bit more time to generate, but you have the runtime here. And the point of the AI studio here is to play around with all of the different tools that Google has available to you because they want you as a developer to understand what's possible, play around with it, and then get ideas of how you can build it into your own app and then ultimately they want you to use the API and spend money on the API within your own app.
Okay, we can click play video and it was an 8-second video that it generated. So, pretty impressive. So, that's the VO generation. You can see I only get 10 free generations of this cuz this is pretty expensive to generate and then I'd have to use the API. Be a little more thoughtful about the videos that you want to generate cuz you are limited for free. So, that's everything here with the generative media that I wanted to go over. You can also see they have like examples that you can click on. So you can see the models that they were the prompts that were used, the models, the system instructions. So this can be a cool way to get inspired of what's possible in terms of generating media.
So next we come to my favorite feature which is building apps with Gemini. So this allows us to build apps that integrate any of these APIs, test them out, and then if we want to continue building, we can sync it with GitHub and then continue building in cursor for example. This one could be cool. It's paint a place, generate a watercolor painting from any Google Maps address. So here we can preview the app. We can see the code. Actually, if you want to save this like what they did as this template, you can use it, sync it to your GitHub, and continue building off of that. You can deploy it. You can share the app. Download the app, download the code, and copy the app if you want.
So, let's say we want to add an address. So, we're going to paint a place. I've just put it on the White House. Let's create a watercolor. So, I guess what it's going to do is take the image from Google Street Maps and then use, I'm guessing, the image in model, but it might be using Nana Banana. I'd have to check to turn it into a watercolor. And okay, pretty cool. It turned that into a watercolor. I like that it signed it Gemini. That's cool as well.
So, there's a lot of app templates here or app showcases that you can see, but then we can also build our own. So, we can decide whatever we want to build. You can start from a template. So, we could say we want it to be use video generation, image generation, chat example, or we can just start from scratch and decide whatever we want to build. So let's say I want to build a video image generation tool that lets me generate images and then ask for changes as well. Let's use the nano banana model. And then there's an advanced settings. You can choose what model by default in only the option is the Gemini 2.5 Pro model for coding cuz that is going to be the best. You can choose the coding language. I'm just going to stick with React as their default. You can add files if you wanted to add like images for design inspiration, but let's just go ahead and run that prompt.
Now, I didn't put a ton of time into thinking what prompt to do. You can obviously put in more effort, put a more detailed prompt, and get even better results. But here, as you can see, it's planning everything out. This is where it gets similar to a tool like Lovable if you've used Lovable before, any of these AI app builders. The nice thing is that it's completely for free. The one thing to note is that of course it's only going to build with the AI functionality of that Google has here in AI Studio, but it's pretty cool that it is all integrated. So it makes testing these like AI features in your app a lot easier and very easy to build in.
So here you can see on the left side we have our chat very similar to what you might be used to with these types of builders. We have our preview it just finished. We have our code and we have the ability to again save to GitHub and you can just connect it to your own GitHub account. So all the code it doesn't have to be it's not here forever. You can take it off of AI Studio.
So the image prompts let's just say a spaceship flying around Earth. So we'll generate an image. And this was all coded just right now. Okay, we have a spaceship now. It it did what I wanted. Now I have the ability to edit it or I can start over. So, I'm going to say let's make it go around Mars instead. So, let's add the image. And it should keep the spaceship looking pretty similar cuz that's something Nano Banana is really good at is maintaining that with edits. Yes. So, you can see it kept that so similar it didn't even change the place. And then now we have Mars. So, now let's say can we add a second spaceship in the background? And we'll edit the image. Cool. Okay. So, now we can see it did it actually added two. Maybe not exactly what we want, but now let's add some features. One thing that's cool, Google has these suggested features right off the bat. Why don't we do that? Add, save, download button. It basically, not only does it have instructions, but it also wrote them out in a more detailed way to make it easier for AI Studio to build it out. So, let's test that and see if it works. Okay, so it says it added it. So, let's try one more time. Spaceship taking off from Earth. So, now let's do slightly different prompt. And we should have a download button now. All right. And yep, we do have a download button. and it just downloaded to my computer successfully. So that worked.
Let's try something a little bit more complicated. So let's say now I want to see the image history as I make edits. Okay. So now let's try it out again and see if it works. So we'll do a spaceship option and let's see what happens. Now I'm going to say make it daytime and edit the image. And now hopefully we're going to be able to see the history of the edits. Okay, cool. So yeah, it built that in. And now you can switch between the history and you can start over as well. And of course, there's a lot more we could build into this, but this hopefully shows you what's possible and how easy it is to build these prototypes of AI apps using any of the AI functionality that we went over in this video.
So, I hope that you found this video helpful. Again, go to ai.dev to build with any of the tools that we went over in this video for free. If you're interested in taking what you've learned and building a full SAS app, check the link in the description to go to our AI SAS course where we teach you how to go from start to finish from idea all the way to a fully functioning SAS app building with cursor including integrations with Stripe for payments for subscription payments everything that you need to have a working SAS app. Click the link in the description. We're giving away the first lesson for free, so you can try it out and see if it's something that you would be interested in. Leave a comment below if there's any questions I can answer, and subscribe to our YouTube channel for more free content like this. Thanks for watching. and we'll see you in the next.