Learn Google AI Studio in 50 minutes

1. Course Intro: In this video course, learn Google Ag Studio and its concepts. Google A Studio is a browser based ID that helps users experiment with generative AI models. With a user friendly interface, Google Ag Studio simplifies the complexities of AI development. Easily write prompts, compare AI models, generate codes, and use the prompt gallery to get the exact prompts directly. Read and scan images and videos with these with that record audio and ask the AI model to get responses. Use the stream real time feature to talk to Gemini, share your webcam and share your screen to get immediate responses and feedback. Tune a model, set the safety settings, get the code, and open it directly with Google Colab. Also enable Google search while writing a prompt and set safety settings for the prompt responses. In this course, we have covered the following lessons with live running examples. Let's start with the first lesson. 2. Google AI Studio - Introduction & Features: In this lesson, we will understand what is Google AI Studio, its introduction and features. With that, we will also see the uses. Let us start. Google AI Studio is a platform provided by Google for developers to build an experiment with AI models. It offers access to the latest AI models, including the Gemini multimodel generative AI models. Let us see the features, you get access to advanced models. Okay. With that, you can quickly integrate AI capabilities into your applications with a Gemini API key. You can customize the default behavior of Google's foundation models. With that, you can easily experiment with those models, easily work around and test prompts. You can also adjust model behavior with techniques like tuning and structured prompts. You can also set safety settings so that you can block specific content. Developers can also find tune models, and whatever you'll create on Google AI studio, you can easily get the code for it. With the real time streaming feature in Google AI studio, you can interact with AI models in real time providing immediate feedback and responses. Let us now see the uses. With Google Eyes Studio, you can write prompts easily because a prompt gallery is provided with multiple examples related to specific task. Also write prompts with specific instructions. Compare the output of LLM models. Get structured output for your prompt. Let's say you want your prompts output in a JSON format, then you can easily achieve this with Google AI Studio. If you want to generate codes, you can easily generate in multiple programming languages. If you want realtime update, you can enable the Google search option so that the output is in real time. If you want to disable harmful content, then a safety setting option is available. With the safety settings, you can adjust how likely the responses could be harmful. Content is blocked based on the probability that is harmful. Some options can be set for this like block low, medium, and high probability of being harmful for harassment, hate, dangerous content, and other harmful responses. Easily upload images and video and use the Google A Studio to read and scan it. With that, you can give audio prompts also. For your projects, easily get codes on the click of a button, and you can also open these codes directly on Google Colab. Tune a model easily based on specific task. With that, an awesome feature is provided by Google A studio to stream in real time, talk to Gemini easily, share your screen, or you can also share your webcam. To get immediate feedback and responses. Share your webcam or your screen to identify objects, to read text. Guys, we saw what is Google E studio. We also saw some of its features and uses. In the next lesson, we will log in Google Ee studio for the first time. 3. Google AI Studio - Login: In this lesson, we will see how we can login to Google A studio for the first time. Let us start. At first, go to the web browser on Google Type, Google Aye studio, press Enter. The official link is visible astudio.google.com. Click on it. We already logged in with our Gmail ID, so directly it is asking for the consent. Click on it. I'll also select this to get updates regarding models. Click IACP. Now here we are inside Google A Studio. 4. Google AI Studio - Settings: In this lesson, learn some important Google AI studio settings. We will set the dark mode for the Google AI studio. With that, we will also enable the autosave option. Let us first change it to dark mode. Here, I'll go below settings. And here, I'll say dark. I have enabled dark mode. Here is the APA plan information. So from here, you can enable autosaving. We will try, but it is not working, so I'll go here and select Enable AutoSave. It will again ask me to login. So I'm logging with my basic Gmail account. You can use any Gmail account. So it enabled autosave. Go here again, settings and you can see auto saving unable now. Okay. Now when you go here, you can see you already logged in with your Gmail ID, click here, and from here, you can switch your account. You can also check the APA plan information from here. So these were the settings 5. Google AI Studio - First Prompt: Now we will see how to write our first prompt on Google A studio. So when you reach a studio, click create prompt like this, first, you can also set the system instructions. It is optional, so I'll do it later. From here, I can try the suggested prompts. I'll write our first prompt. Suggest me a recipe for an Indian dish, high in protein. I'll click Run. So this is our first prompt. Okay, so it is giving us an answer. From here, you can clear the chat. Okay, go above and you can see the options. Cena Masala. Click here and Ron, you can edit it. You can also click more options and delete. You can also copy the text or you can also copy the markdown. This was our first prompt. We did it with the default settings here. Here we can also edit the title and description. Recipe suggestions, save. Okay, so we saw how we can write our first prompt. Here is the prompt. We already enabled Auto save, and it got save to drive on its own. 6. Prompt with system instructions: In this lesson, we will see how we can set the optional system instructions while writing a prompt using Google A Studio. So this is a great feature. So in different AH ad boots, we saw that we need to mention in the prompt itself that let's say I'm 10-years-old, I'm 20-year-old and I want to learn something. Okay, we can set it here using the system instructions option. This is optional. You can set the tone and style instructions for the model. So let us create a new prompt and try the system instructions. Click Great prompt. And here I'll set explain me like a 10-year-old. What is Python? Simple prompt, what is Python? Click Run and the section is Explain me like a 10-year-old. Click Run. So it is explaining me. If I did not mention this, I don't think it will give me this line also since it is explaining like a kid. Here I'll edit the title and description. A prompt with system instructions. You can also add a description. I'll click Save. So this is our second prompt, and it got saved here. We have two prompts now. It is visible here. All the prompts are visible here. Okay. So, guys, we saw how we can use system instructions while typing a prompt. 7. Select the LLM i.e. the model: This lesson, we will see how we can select LLM models before writing a prompt in Google A studio. So we wrote two prompts till now, and here are the settings, run settings. The model, the current data showing Gemini two point oh flash, when you keep the mouse cursor, all the details will be visible. Like here, you can see the context length, its use case, its features, the rate limit, also the latency. I'll click here, and different models are visible. You can see keep the mouse cursor and get all the info. Go below, and these are the preview ones, also the gemini, also the Gemma. Okay. I'll keep it as it is. 8. Compare the LLM model's response: In this lesson, we will see how we can compare the models. Till now we wrote prompts using the default model, which was Demini 2.0 flash. So now we will compare the models. I already clicked on Create Prompt. I'll click Compare and you can see different models, select any model. I'll select flashlight preview here. Now let us type a prompt and it will give a result for both the models. AI is a boon or a BAN it will show me the output. Parallel, click Run. Here it is you can see. We will also check the speed. Here you can see. Okay, 6.4 seconds, 4.5 seconds because the left one give a bigger answer. You can also click here. 762 tokens. 749 tokens. Okay. I'll click here, Edit. Compare models or compare LLMs. Click Save. So this is how we can compare different models. 9. Token Count: In this lesson, we will understand the model parameters in Google studio. So there are different parameters available that you can set before writing a prompt, like token count, temperature, and others. Let us see I'll create a new prompt. Okay. Now you can first select the model I already told you. We will keep it the same. Then comes your token count. It counts how many tokens are used for content generation. The following is the limit right now, 1 million so here, token count refers to the number of individual units that a piece of text is broken down into by a tokenizer. It can be words, sub words, or even individual characters. A models often have a maximum token limit for the input text. So this is the limit here right now. So this will automatically count the number of tokens. We just printed up prompt and it was visible here that how many prompts were there. Okay, we will do it again in this lesson. I can type here. Let's say this, who is Vatico. Can you see the token count? It will change. Okay. It's 237 now, so it can be individual words, characters. Okay, and the limit is here. This was about token count. 10. Temperature: Now comes temperature. What is a temperature as a model parameter? In EI models, temperature is a parameter that controls the randomness of the model's output. It is particularly relevant in language models where it influences how the model select the next word or token in a sequence. So how it works, it works like this. Okay, you can allow the creativity in the responses by dragging this here and there. Okay, I'll keep the default one only. Low temperature results in more predictable output, whereas high temperature encourages more randomness and creativity. So I'll keep the default. So this was about temperature. Okay. Let me set it to the last two, and now I'll ask who is Roy Sharma, click Run. Token count is visible and the temperature is two now. Let's go below. So higher temperature means randomness and creativity. That means more creativity. It's written creativity allowed in the responses. You can compare both of them above. Okay. It added more subpoints here. So, guys, we saw what is a temperature in AI models. We also saw how to test it. Thank you for watching the video. 11. Structured Output: In this lesson, we will see some of the tools available in Google A Studio. Okay, so let us begin with the structured output. Here it is. I'll click on Create prompt. Now we will work around the structured output. In Google a studios structured output refers to the models response being formatted in a structured data format such as JSON rather than unstructured text. This is useful when you need the output to be easily passed and processed by other systems or applications. Okay, so here is the structured output. First, you can click here. Now we can mention a prompt. I have a prompt here regarding erat Cool. Okay, I'll mention this and the output will be structured in a JSON format. Okay, it's about his contribution towards cricket in different championships. I'll click Run and let's see what will happen. So here, can you see it has shown us a structured output? Okay. In a JSON format. So this is how you can convert your prompts output in a JSON format. That is a structured data format. It is useful when you need the output to be easily passed and processed by other systems or applications. Through this you can easily extract information from text and categorize content. It has mentioned the teams and the achievements of Varadkl in JSON format. So VariatKl is a cricketer from India. So you can also download the JSON format now, also copy it from here. With that, you can use more options and copy the text or even the markdown and you can also edit. Guys, we saw what is structured output in Google Studio. Thank you for watching the video. 12. Code Execution: In this lesson, we will see what is code execution in Google A Studio. So you can generate codes easily on the click of a button with Google A Studio. Let us see. Okay, so under run settings, you have this code execution. I'll just uncheck this and check code execution. Okay. So you can execute the code in the prompt. I'll click Create prompt. Now, let's say I want I want the 30th prime number. So I'll just click Run Okay, I'll select code execution again. And I'll click Run. Okay, so it has written the code for us. Here is the complete code to calculate the 30th prime number. It has given the output also. You can copy the code from here, you can download the code. With that, you can also download the output and copy the output from here. So this is how you can generate codes with Google AI Studio. It has also mentioned the language here. We could have also mentioned in which language we want, but we kept the default. 13. Prompting with Google Search: In this lesson, we will see how we can use the grounding with Google search to get the recent results. Let us see. So here under settings, you can see grounding with Google Search. So what I'll do first, I'll write a new prompt. Okay, and then I'll use the same prompt after enabling this option, grounding with Google Search. Okay. I'll type when and where is the champions trophy? Without enabling this, click Run. Here it is. It is showing that the champions trophy will held in 2029, which is incorrect. Okay, because we did not enable Google Search. Now I enable this, and I'll copy this prompt and paste it here. Click Run. It is by Google Search. Obviously, both the products are by Google. So here it is. The champions trophy held from February 19 to March 9, 2025. Okay, and these are the sources. Also the Google Search suggestions. So it is powered by Google Search. So this is an amazing feature. Okay, so you can get the real time of dates also. So, guys, we saw how we can use the grounding with Google Search and use Google search results for our prompts. Thank you for watching the video. 14. Safety Settings: In this lesson, we will see how we can use the advanced settings in Google A studio. Let us begin with the safety settings. First, I'll create a new prompt and on the right, go below, advanced settings, I clicked on it, and the safety settings are visible. I'll click Edit safety settings. Okay. This settings help developers manage and mitigate potential risk associated with AI generated content. As written adjust harmful response settings. Okay, so you can adjust the level of safety filters and content restrictions based on your specific use case. I'll click here and you can set Always show regardless of probability of being harmful. You can also tackle hate speech issue here with that block dangerous content also here, I just want to show you something in the previous prompt. Can you see this? So this was something which I wanted to discuss the safety settings. Right now, it's low. We can also adjust safety settings from here, Edit safety settings, and it's low right now you can see. So we can adjust it. We can set it, block few, block some, block most, asmment, okay. Block civic integrity block. You can set it from here or you can also set it to medium. I have enabled it here. Be clicking here, and we now have the safety setting, so it won't show any harmful content. In this way, guys, we can work around the safety settings on Google Eye Studio. 15. Add Stop Sequence: In this lesson, we will see how to work with the stop sequence settings in Google A Studio. With that, we will also see an example. Here is our Google A studio. I clicked on the create prompt. Okay. Under advanced settings here on the right, you can see stop sequence. So let's say I want to set a specific word or a character so that the model stops generating text when it encounters this sequence. Okay, so a specific sequence of tokens that when encountered signals the model to stop generating further output is what we call stop sequence. So here I'll set five. Let's say for example, and now I'll type a prom give me the top seven places to visit in Australia. It will now generate. Here you can see it did not generate it seven because I have added a stop sequence. So when it encountered five, it stopped. So this is the purpose of stop sequence. So in this way, you can limit the length of the generated text by enabling the default settings. So guys, we saw what is a stop sequence in Google A studio. Thank you for watching the video. 16. Set Output Length: In this lesson, we will see how to set the output length in Google A studio. Let us see. So as the name suggests, the maximum number of tokens you want in a response is what we call output length. So first, I'll disable this. Okay, we already understood what is a stop sequence. Click create prompt. Now I'll ask Okay, generate a fitness plan for 25-year-old guy. Okay, it has automatically set the maximum number of tokens. So I'll just click Run. So it is giving us all the details related to the fitness plan. You can click Stop here, but I'll keep it as it is. You can see it is not stopping because the output length is 8192 tokens, maximum. Okay. Now I'll again take this prompt. Okay. Here I'll set 200. The maximum number of tokens. Click Run. We have set the maximum number of tokens in response. It's too short right. Okay, maximum number of tokens are now 200. So this is how you can work around the output length feature to specify the maximum number of tokens the model can generate for a given response. So through this, you can limit the length of the generated text. That is the output. That is response. 17. Read and Scan Images: In this lesson, we will see how we can read and scan images with Google Age Studio. Let us see. This is our Google Age Studio, click Create prompt. Now here an option is visible, insert images, videos, folders, files or audio. Okay, so I'll click here plus you can select from your Google Drive, or you can upload a file. You can also record an audio or enable the camera directly, or you can get a sample media from here. I'll click Upload File. I'll click here. What is this image about? Click Run. Let's see what will happen. This is a thumbnail for a Video tutorial about a free database. It read the text completely, and it is large and bold letters, and a database icon is also visible. The text is Video description. Okay, so it read the image. So, guys, we saw how we can easily scan images with Google A studio. 18. Scan videos: In this lesson, we will see how we can upload videos to Google AI studio and scan it so that we can see how Google AI Studio works when we upload videos. So let us start. Okay, here, I'll click Create Prompt. Now, let's say I'll click Plus here. I already told you this will allow you to upload. I'll select Upload File. Okay, so here it is. One of my older videos, I'll select this, click Open. Okay, so here it is. What is this video about? I'll just try to get some info about the video. So this is a tutorial related to Android Studio. Here it's written, it's about Android Studio. It is one of my older videos. Okay, and it has given us the complete description by scanning the video. So, guys, we saw how we can scan videos with Google AI Studio. Thank you for watching the video. 19. Use Sample Media: In this lesson, we will see how we can work around the sample media provided by Google studio to scan and read it. So let us see I clicked Great prompt on a studio. Here, I'll click the plus sign. The following is the sample media. I'll click and the sample videos as well as the media is visible. So what I'll do, I'll just click on the following and click at to prompt. Can you suggest similar images? What is this image about? Click Run. Okay, so it read the image completely. Astronaut in a space suit with the suggestions. Okay, so this is how we can work around the sample media. So, guys, we saw how we can get the sample media video images provided by Google A Studio itself. Thank you for watching the video. 20. Record Audio and ask: In this lesson, we will learn how we can record an audio on Google Eye Studio. So let us see. Let us create a new prompt, and here I'll click plus record audio. So I just need to click on Start Recording. So I'll ask something. Start recording. Allow this time. Start recording. I'll speak now. Hi, I'm from Delhi, India. Can you suggest some places to visit in Delhi for my friend? Thank you. Click at to prompt. Okay, recorded the audio. You can play it from here. Click Run. Now, let's see what will happen. It has also shown us the tokens. Here are a few places to visit in Delhi, India gate Kotumna. So the audio stuff is also working. In this lesson, we saw how to record audio and give voice instructions for a prompt in Google AI studio. 21. Get code: In this lesson, we will see how we can get the code. So whatever we did till now, we can easily get its code. Also, we can get the API key and walk around. Let us see. This was the last prop we recorded an audio and got an output from the Google A studio. Now you can go to the run settings and click Get Code. Get template APA code for your prompt in different languages. I click Get Code, and you can create your APIKey before using the code in your project. When I'll go below. Here you can see the Gemini API key, you can click here to get the APA key and rest of the code is here. Okay, I, what I did, I have used the model Jamini two point oh. You can change the model also from here, from here or you can directly change it. You can copy the code from here. Also, you can get the API key from here. By clicking get APIKe. Create APK from here. Okay. So in this way, guys, using the Get code feature, you can get the code easily. Also, you can change the language from here. You can get in Python, JavaScript Go android Okay. So, guys, we saw how we can get the code in Google A studio to get the APK, just click here and create an APK. After getting the key, place it here. Okay, thank you for watching the video. 22. Open code in Google Colab: In this lesson, we will see how we can open our project code in Google Colab directly, so there won't be any need to copy the code or work around it. We can directly open it in Google Colab. Google Colab is a web application provided by Google to run your programs. Using Google Colab, you can combine executable code and rich text in a single document. So let us see how we can open it in Google Colab. So this was our last project. We added an input in real time. That is by recording an audio, and we got the output. Click Get Code. We already saw the Get Code feature. We can get in any language. So here we have this open Colab featured. Directly, you can open it. Okay, I'll click. Both Colab and a studio belongs to Google. You can directly create a key from here. By clicking create your first key. So I've shown this in the previous lesson also. Click create APIG Okay, so here it is. Click here and you can copy the APAKey from here. Okay, remember, you don't need to share your APAKeyO never use it publicly. Therefore, it is blur here. Click Copy. So in this guys, you can easily open your code in Google Colab. With that, you can also easily get an APAkey for free by Gemini API. Thank you for watching the video. 23. Tune a Model: In this lesson, we will see how we can tune a model with Google A studio. Let us see. So under a studio, click Tune a Model. Here you can select the data for tuning. You can select an existing structured prompt or you can also create one. If you have a Google Sheet or a CSV file, you can easily use it. Okay, tuning a model in Google AI Studio involves fine tuning the model to improve its performance on specific task. You can also prepare your dataset. The dataset should include input output pairs. Import your dataset into Google A Studio. Like you can select the data source from here or you can import. When you click Import, it will directly take you to Google Sheet. You can add your model name with the description, also choose the base model, and under advanced setting set the parameters. Here you can set the epoch. That is the number of times the model processes each example in the training data ones. With the multiplier and the bat size, learning rate multiplier is a factor to adjust the default learning rate for the model. The bad size, consider it as the number of examples processed at each iteration. Okay. Let's say we will select a data source news headlines. And here you can see the sample. Here it is only showing one input and one output. Here is the input regarding BBC News. In this way, you can work around and tune a model. Okay, so let's say I'll keep it as it is. I'll go below. Okay, I'll name it first tuned model and click Tune. Here it is showing. It will take some minutes. You can end a tuning also. Let's wait. So you can see it be completed. Okay. Okay, it should be first tune model. I clicked on it to Cedar Details. Here is the bat size and other details and the performance here. You can directly use your tune model at APXs, I'll click Unchat. Okay, now you can see it automatically selected the first tune model, and it was here. So your tuned models will be here. Okay. When you'll go for the model. I can click here. I'll just add a correct name. First tune model, you can add any name. Click Save. Now, it's fine. Okay. Okay, you can see, using chat. Now our model is based on what we created. That is the sample. So, guys, we saw how we can tune a model in Google A studio. Thank you for watching the video. 24. Prompt Gallery: In this lesson, we will see what is a prompt Gallery in Google A studio. Let us start. So when you open Google A studio here, prompt gallery is visible. Click on it. Now you can see some great prompt ideas are visible. You can directly go for the math, Tutor, solve geometry problems. Check which number is bigger, optimize the code. Okay. You can also check the sentiment. I'll click. Okay, so it's written, positive or negative. It's so cold today, I can't feel my feet. It's negative. The weather today is perfectly adequate. I'll type something. I'm a myth, and I love cricket. Cricket is live. Click, Run. It's positive. Okay, so this is how you can use the prompt gallery. I'll click here. Let's check this, which is bigger, which number is bigger? What is bigger nine or 9.11? Run. 9.9 is bigger because 9.9 it considered it as 9.90, so it is bigger than 9.11. So in this way, you can work around the prompt galleries, play around with this. So, guys, we saw what is prompt Gallery in Google Eye Studio. Thank you for watching the video. 25. Stream Realtime: Talk to Gemini : In this lesson, we will see what is stream real time feature in Google Eye studio. Let's see. So with the real time streaming feature in Google Aye studio, you can interact with AI models in real time providing immediate feedback and responses. Let us see. So this is our Google Aye Studio. Click stream Real Time. With this, you can talk to Gemini, show your webcam to Gemini. Okay. Also you can share your screen. With that, you can select the model. Okay, with the output format, audio or text, you can also change the voice from here. So what I'll do, I'll talk to Gemini. I'll click. Can you hear me? Yes, I can hear you. Is there anything you need help with? Yeah, so I am from Delhi. So how is the weather in January In Delhi, can you tell me? I understand you're asking about the weather in Delhi in January. Unfortunately, I don't have access to current weather information. So can I enable the grounding with Google Search feature so that you can check it from Google? I understand you're asking if I can use Google Search. Unfortunately, I don't have the ability to access external websites or search engines. Okay, so I think the grounding with Google Search feature is not available in Streamline. Not a problem. So I'll ask something else. Do you know who is Virat Cool? Yes, I know who Virat Cole is. He is an Indian international cricketer and a former captain of the Indian national team. Yeah, you're right. I'm a huge fan of Virat Cool. He's my idol. That's great. You are a fan of Virat Cool. Is there anything else I can help you with? Nothing, so, thanks a lot. You're welcome. Have a great day. Okay, bye. Bye. Okay, so this was the streamline feature. You can see the live. You can disconnect by clicking here. You can also start a new stream. So I'll click here again, or I'll click Great Prompt. Okay. Then I'll click stream Realtime and you can share your screen also. Show your webcam also to get real time feedback. 26. Stream Realtime: Show your Webcam: In this lesson, learn how to show the webcam in Google E studio. This is used to identify objects, read the text. Let us see. Click on Show Gemini. Allow This time. Okay. I understand that you are asking for permission to do something at this time, is that correct? Yeah. So can you see what is this? Can you guess what is this? Yes, it appears to be a bobblehead figurine. The name on the base says Virat Cole, who is an Indian cricketer. I told you I am his fan, right? I hope you remember. You're a fan of Virat Cole? Do you collect many things related to him? Yeah. Thank you for guessing. I was just testing. Ah, okay. You were just testing me. Is there anything else I can help you with today? Nothing. Thanks a lot. Okay, you're welcome. Have a great day. In this lesson, we saw how to show the webcam to identify objects with Google AI Studio. 27. Stream Realtime: Share your Screen: In this lesson, learn how to share your laptop screen and ask Google AI Studio to identify the OS and read text. Let us see. Guys, we saw how we can use the stream real time feature. Okay, so we were able to work around a solution so that we can talk with Gemini. We used our webcam also to display something. We can also share our screen. Okay, so that it can read. So let me do it. It is asking entire screen. Okay. I'll click here Share. Okay. It looks like it's asking for the entire screen or stream. Is there anything else I can help with? So I'll just minimize this and you can guess what is this? Which OS is this? Okay, it appears that the operating system on the screen is Windows 11, is that correct? Yeah, correct. And what is the time right now and the date? Okay. According to the bottom right corner of the screen, the time is 17:40, and the date is 14 February 2025. Correct. So what is the folder name on the desktop? Can you see? Okay, the folder on the desktop is named Amrita. It's named Amit. Ah, apologies. The folder name is actually Amit, not Amrita. Amrita is my sister. Thanks for letting me know. We are twins. You're welcome. Is there anything else I can assist with? Nothing. Thanks a lot. You're welcome. Have a great day. Okay, so I'll disable it. Stop sharing. Okay, bye. Okay, the screen sharing has been stopped. Is there anything else I can help you with today? Nothing. I'm just canceling the streaming. Thank you for your time. Okay. You're welcome. Have a great day. So we tested all the three features of stream real time, talk to Gemini, show Gemini, share your screen. We can also change the voice, the format, also. It can also show the text. Thank you for watching the video.