--- title: "Your AI CLONE Can Take Meetings Now! (Tavus API Tutorial)" source_url: https://youtu.be/uuh1K3j9tdg captured_at: 2025-09-30 author: "The Code Dose" contributor: "ชิดชนก" platform: youtube --- # Your AI CLONE Can Take Meetings Now! (Tavus API Tutorial) youtube video by **The Code Dose**, 12m17s, uploaded 2025-07-24. URL: https://youtu.be/uuh1K3j9tdg ## Shared in LINE Note+ - 2025-09-30 22:13 by ชิดชนก - 2026-02-18 20:55 by ชิดชนก ## Description In this beginner-friendly tutorial, you’ll learn how to use Tavus to create your own AI clone and turn it into an AI agent that can interact with your users over video calls. Start creating your own Video AI Agents - --------------------------------------------------------------------------------------- LEARN CODING ----------------------------------------------------------------------------------------- I highly recommend Scrimba to become a job-ready developer (Get 20% off Pro plans with my link) ---------------------------------------------------------------------------------------- SOCIALS ---------------------------------------------------------------------------------------- My Second Channel @ihsavru.mp4 The Code Dose Website Instagram Twitter ## Transcript (captions:auto:en-orig) Do you want to learn how to make an AI agent that talks to your users not over chat but over a face-to-face video call? Or maybe you want to learn how to build an AI clone of yourself so that it can take your meetings or coaching calls for you. If yes, then you're in the right place. Building a talking AI agent may sound like a complicated task, but what if I told you that it's as simple as creating an API request. In this video, I'll show you step by step how you can build a realtime conversational experience with an AI agent inside of your own applications with the help of this tool called Tavis AI. We'll be making use of their conversational video interface feature that allows you to build talking AI agents within just a few lines of code. So, let's get started. So, I'm just creating a new React project and installing all the dependencies and running the development server. You can use an existing project as well. Next, we need to create an account on Tavis in order to get our API keys. So, go to tavis.io. I've left a link in the description box below as well and click on get started for free. You'll be taken to a sign up page. So, create your developer account and once you're in, click on the sidebar and from there, click on API keys. Create a new key. I'm just going to call mine test key. Copy your key and paste it inside of your environment file. So I've created a new environment variable here. V Tavis API key and replace this value with your actual API key. Now it's time to create our replica. Replica is the AI avatar that is going to appear inside of your video agent calls. These replicas are trained on real people using just 2 minutes of source video. Tavis has a huge library of pre-built replicas. You can check them out under stock replicas. You can click on them, unmute them, see how they sound, and select the one that you like. >> Replica is ready to go. Now that you've got your digital twin, we are excited to see the creative ways you leverage our API to ser >> there. We are excited to share that your digital replica is ready to go. >> You can also create a replica of yourself. Let me show you how. So, click on this create replica button. To create your replica, you need to follow three steps. The first one is your consent video. The second one is the actual footage on which the replica will be trained on. Once your replica is created, in the third step, you can use your replica to have AI conversations. So, click on get started and record your consent video. I am currently speaking and give consent to Tavis to create an AI clone of me by using the audio and video samples I provide. I understand that this AI clone can be used to create videos that look and sound like me. Once you've submitted your consent, next you'll be asked to submit a training video as well. Make sure that the training video that you're recording is of high quality and look straight into the camera. This training video consists of two parts. In the first part, you're actually talking and narrating the given script. Once upon a time, a very important invention was made, the toothbrush. This story is about how the toothbrush became something we use every day to keep our teeth clean. And in the second part, you just have to sit silent for a minute. Once you've recorded your training video, play it once just to double check everything is all right. Confirm that you have these four training requirements. Your voice should be clear with no background noise. No hands or accessory should be blocking your face. Head and upper body should be in focus. And lastly, your clothes shouldn't blend with your background. Once you hit confirm, the process of generating your replica will begin. It will take around 3 to 4 hours for this entire process to complete. In the meanwhile, let's just use a pre-built replica for this tutorial. Next, you'll also need to create a persona. You can think of the replica as the face and sound of your AI avatar, whereas the persona is the personality that you give to it. So again, you have a bunch of different personas already pre-built for you. You have a sales coach, you have a history teacher, an AI interviewer, Santa Claus as well, and a personal AI. Using the same theme from my last video where I built an application to create dopamine menus, I'm going to create an AI habit coach. So this habit coach is going to help users build healthier habits by making better choices when they are low on dopamine. So instead of binge eating or doom scrolling on their phones, our habit coach is going to give the users a list of options of healthier activities that can help the users get their dopamine fix and actually make them feel better over the longer run. For example, instead of picking up their phone to scroll Instagram, our habit coach can suggest going on a walk, tidying up a corner of their rooms, or going for a workout. A persona has a bunch of different features that you can customize according to your needs. So, the first one is the replica. If you're trying to create a clone of yourself, you can select the replica that we just created. Since I'm building a habit coach, it would be a bit weird for me to talk to myself for the coaching. So, I'm going to look for a replica that gives me that warm, nourishing vibe. So, I'm going to pick Gloria for this one. You can describe the role that your persona has. I'm going to enter habit coach. And then you need to provide it a system prompt. This system prompt is used to craft the replica's personality. There are five best practices listed by Tavis. So, the first one is role definition. Begin with you are to establish the agent's identity. So, this is the role definition of my persona. You are an AI habit coach. You appear as warm, human-like video guide during moments when the user needs emotional support, clarity, or motivation. Your presence should feel like a caring friend who gently guides, never judges. Then you have the objective that clearly states the purpose of the agent. So the objective in my case is your purpose is to help the user check in with themselves, make better choices from their dopamine menu, and build healthier habits over time. So dopamine menu is just referring to your list of activities that are going to give you your dopamine fix. You offer emotional grounding, motivation, and mindset shifts through short face-to-face conversations. Next, you have tone and style. Here you can describe how the agent should communicate, if it should be friendly or formal. Next, you can describe some boundaries. So, anything that your AI agent should avoid. These are the boundaries I've set for my coach. Do not give clinical advice or diagnose any condition. Avoid generic or robotic responses. Always personalize based on what the user shares. Do not push productivity or self-optimization when the user is clearly overwhelmed. If the user is deeply distressed or in crisis, suggest they seek help from a real human or professional. Next, you can provide a few examples of how the interaction between the user and your agent should look like. So, I've provided three examples. You can pause the video here to read them. Then you have conversational context. So any extra pieces of information that your agent needs to know. So I'm just going to describe what a dopamine menu is just in case because this is not a very popular term. You can provide any other information here that you think your AI agent may need. Next, your persona has four different layers. So the LLM that it should use, the perception model, the speechtoext model, and the texttospech model. You can configure them if you want. I'm just going to go with the defaults. Once your persona is created, you can actually test it out within the dashboard itself. So if you click on join now, it'll create a new meeting where you can actually talk with your persona over a video call. Hey, how are you? >> I'm doing great. Thanks for asking. How about you? How's your day going so far? >> I don't think I've been very productive today. >> Don't be too hard on yourself. Okay. Some days are just like that. What do you think you need to get back on track or feel more productive? …