Transcription
All right, we are live. Good morning, good afternoon, good evening, wherever you're joining us from in the world. I'm just going to check that we are streaming and we are good to go. Yep. All right, we are super excited this morning. We have live avatar deep dive and we have someone new for the first time on a community webinar. Sunny is joining us and we have Chloe who if you've come to our previous interactive avatar walkthroughs or deep dives, you know Chloe. Um, so please give Sunny a warm welcome in the chat. We're going to get into the deep dive. Um, but first, if you are tuning in on YouTube, you're always welcome to come join us in the Hen community at community.hen.com. That's where we'll take Q&A at the end of this session. So, if you're watching the recording, too, come join the community so you can join us live next time. You can find tons of resources on the community. Um, Chloe and Sunny will walk you through resources for Live Avatar by Hey Jen today. Um, and I don't want to take too much more time when I hand it over them. They've got a lot of great content. If you are tuning in and you are new to Hey Gen. A great place to start is the community, you can check out Hey Jen Academy for intro sessions. And today's session is focusing on live avatar. If you're also interested in Hey Jen's video creation, you can come find out more resources, dive into Hey Genen, and explore that. Um, but today we're going to dive into Live Avatar, which is a very exciting part of Hen's business. So, with that, I'm going to hand it over to Sunny and Chloe for today's session.
>> Thank you. Thank you, Audrey. Well, welcome everyone and it's such a pleasure to be on this webinar and excited to share with you our new modalities here at HEN called Live Avatar. Today we're going to dive a little bit deeper into the business applications of this technology and share some great customer stories and also have Chloe to walk us through about some technical integrations and how do we connect. Um, shall we kick it off now Audrey?
>> Yeah, let's jump into it.
>> Okay, great. I'm going to walk through the agenda really quick and this is something we're going to cover today. Uh, the first, I'm going to talk about what is live avatar exactly and how do we distinguish ourselves uh among other streaming avatars and also real-time avatars as well. And the second, we'll talk about the timing, why do we bring it up now and why is it important right now and what are the business applications and use cases that could be relevant at the moment. And then we're going to share some successful customer stories and really quick, just walk through them. I'm going to show you a very exciting demo video um and also a chat demo with some of our live avatars. And then by the end of the sessions, we'll have Chloe to show you some of our standard API and SDK and how do you can how can you connect and use for your own. Okay, so let's dive in. Um, so the first pillar is called what is live avatar? So, live avatar was called interactive avatar two years ago. It was a product at our labs um and where we wanted to develop and research about this technology and make it a viable and stable. And two years after, we feel like, okay, it's ready and we take it out of from lab and make it a real product and um invest it and invest a lot of resources at HEN to make sure we serve our customers really well. Um, so what has been special for Hun's interactive avatar that we give them a new branding called live avatar is that something we're extremely proud of is Hun's ability to produce the hyper-realistic AI avatars that we focus on so much effort in our model training to capture those facial expressions and make sure we have this real, real person feelings when we talk to the avatar, which come down to build this great foundation to enable this technology where people can talk to the avatar and feel it's real. And the second is that we absolutely noticed that um, customizations of the avatar, it's so important. So, some of the people just need an avatar there for their commercial use case. Some people uh need a specific person to be the avatar. You can do both and we provide that flexibility to you. Um, so the first, like you can actually for the first use case, you can actually just choose any avatars um in our public library or you can u film a two-minute footage of your chosen person and upload it and boom, and you have your own avatar. So that customizations is something that we are proud of as well. And the third is latency. As we come down to serve the commercial values, latency has been so tremendously important, right? So that's why we put a lot of engineering effort to make sure we improve latency and we can serve the commercial use cases really well. And also from the infra level, that uh, we wanted to provide a very like state-of-the-art industry-leading SLA. And those are the efforts that we put into this technology and product and proudly to make sure it is good enough to be embraced another of like huge growth at Hijen.
So, and now I'm going to talk about timing. Um, so for the past two years, I'll say one year and a half, we've been talking with a lot of customers. Um, and somehow we feel like the timing is a bit early. Um, it is not yet there and the market is not really ready. Um, now we feel it's definitely the timing to bring it out because over the past month, we definitely sensed the high demand almost from all industries that customers really want that human touch, human experience and to build the connections with them and their businesses they they like. So that's why I think live off track can serve a really good function there. And and then scaling is also another problem as many of our customers are they are global businesses, especially when they're in um in tech industry where their product and audiences is usually for global market. Um, some of them are startups. They don't even have the ability to to run a large organization if they wanted to scale. So live avatar comes in as a quick solution for that. Um, so to help them to serve as many customers as we want. And the third is about 24/7 operations. As I mentioned, if you wanted to target a global market, then you you are covering like different time zones, right? So think about if you are operating eight hours a day and your competitor is operating 24 hours a day. Well, guess what? You're out. So that's something the competition grows and somehow your business has to evolve and match that capability to be able to compete in the market. So I think that competition is really getting intense right now. So 24/7 operation for most of the businesses is pretty much a must-have, especially in the customer services or customer support scenarios.
All right. So the third pillar I'm going to cover a little bit is about the use case and this is the four main use cases that we already seen a lot of success among our customers and we're proudly to say, okay, this really works. This is not something like fake PMF or something like that. It is that we really bring tremendous commercial value to our customers in these four sectors. So e-learning assessment, I think this is the first one which is very easy to understand. Like teachers can talk to the students anytime like avatar teachers, right? Students like they don't have to be a full session like 40 minutes full session. They can talk to the avatars anytime to practice the knowledge and also the syllabus. Um, and the second is the AI recruitment where avatars bring the interviewer for the recruiting company or the AI department of a certain organization to do the first two screening or the first screenings and just, you know, capture the behavior interview that we all think through and capture those informations for the employers and then do analysis and see which one should go to the next round. And this is something we feel we see tremendous success as well. Um, in sales trainings, sales training is a really interesting use case in my opinion because you definitely notice that for every almost every business person that when you get into a new field, when you join a new company, you need to practice the knowledge about the product and industry. You need someone to to practice against. And for negotiation, for objection handling, and like 24/7 in different time zones and multilingual environment, that is how much you can enable your sales training with live avatar. The fourth is learning and development. Um, this is a very interesting sector because um, some of the classic use cases in sector is for example, if you want to do an internal training for a fast-evolving industry or fast-evolving product, like you need to keep updating the knowledge base, you could updating the new west informations about certain things and do the internal training for your employees. So that is somehow you can just update the knowledge base for the live avatar, then it can have the real-time like Q&A with with your internal employees with the newest knowledge. So that is something really fit into uh the industries who are fast evolving. So those are some of the use cases.
Now, I wanted to show our newest website to all of you and we can just walk through you guys about some of our success customer stories. Let me stop screen sharing and reshare it with our different tab. One second.
>> [snorts]
>> Can everybody see my screen? Great. Thanks, Chloe. Um, so this is our newest website. Uh, we just launched it last week, last month. It's very exciting. Um, so you can see there is some of the demo avatars and you can talk to them. But now I wanted to dive directly into the use case and share some successful customer stories with you and to show you the real numbers and the real data out of this technology. So let's dive deep into this. How ADU can act. Edu is actually one of our customers in the e-education sector. Um, what they do is that they collaborate with institutions and universities and to help the students practice the syllabus and so to improve their test scores. Um, so after using live avatar, they basically closed every prospect that they wanted to close. So people are so wowed about this technology and a lot of their customers loved it. And we see a very concrete increase of 15% of uh students' test scores with only 15 minutes per day interaction with the lab avatar. And that is absolutely amazing. And because of that, and you have a 3,000% growth, like the company literally booming after uh um applying this technology. And I think the 24/7 is very obvious, right? And so that is enabled edu to be able to capture the global market and work with uh companies in different time zones. And so I'm going to quickly just walk through because we have a lot of agendas afterwards. And the second is recruiting. Um, we work with recruit company Zeligate. Um, so we saved them 24 hours per hire because most of the avatars can conduct the first of two screening interviews for uh Zelgate and for their clients. And the avatar was able to do 14,000 interviews for in a very shorter period of time, which enables delegate incredible scalability. And they don't need to hire anybody new to do that, which is a tremendous success. And that's something I'm personally very excited about is that most of the audience and interviewees report that they feel more comfortable talking to an avatar than talking to a real human being because the avatar is unbiased. They don't feel being judged, especially for people who in color or where English is not their native native language, they feel like, okay, the [laughter] avatar is not going to judge me, so they feel more comfortable, more confident. So the customer satisfaction is definitely increased. Um, and because of that, delegate was able to save some cost of hiring new interviewers. Um, so the company actually actually grew 35% over time. Um, this is something I feel like it's a great use case to share um in the recruitment. [clears throat] And the third thing is language learning. I think this is very straightforward um because language is something that you definitely need practice with, right? If you're learning a new language, it's very likely you don't have anyone to practice with, right? And you just pull out the avatar and trying to practice it like two minutes a day, five minutes a day, but at least who is patient, who is professional and fluent in those languages and you have this comfortable feeling that, okay, this person is not going to be um, you know, put off by your very poor pronunciation. Um, so uh, Speakology worked with 600 schools and to apply live avatar to help students to practice the language. Um, and they investigated a 10 times more um student engagement in different sectors. So this is something that we feel extremely powerful when it comes down to language practice and education. This is sort of an education pillar as well. And the last but not the least is the sales training thing. Sales training um like I said, we need someone to practice our negotiation skills and and our um objection handling. We need to feel comfortable about any business scenarios. And so this is not just for business people, right? This is almost for everyone. So as long as you're in the business, you are in the industry, there are some sort of scenarios you have to deal with, right? So the internal training um use cases for this sales training use case is actually very common to see um and we work with copi and AI and sets the curve for over 2,500 hours of sales trainings only conducted by live avatar. Think about that. And before this technology, what kind of resources you have to invest to be able to do this. So yeah, those use cases are all available on our website if you want to dive deeper into those use cases. We have very detailed articles and just stating how the customers are using our technology and and and what are the fun stories behind.
Okay, cool. That's enough of talk from me. Um, so now I'm going to show you guys um a demo video, a launch video of live avatar where we include all the stories of these customers and how do they feel after collaborating with us. Let's go.
>> Hi, I'm Reed AI Reed Hoffman's digital twin powered by Hen's Avatar Tech based on 20 years of Reed talking and making content. You know what's weird? That this is because of you and [music] your company. You birthed me, mommy.
>> Oh my god.
>> So, we were looking for a solution that would give us this realistic, you know, humanlike approach. Live Avatar and HJM was by far the best quality we could find.
>> Live Avatar as a product has been exceptional because it's programmable, deployable, scalable, [music] and reliable.
>> So, I said, "Let me find the best in the business." And I just found that your company has the best workflow, the best product, and the best customer service.
>> Can you say hi to the group here in Manoring?
>> You might have a completely knowledgeable expert chatbot, but it will never have the presence any avatar would have with an end [music] user.
>> We cannot interview 150 candidates. Now with the live avatar, every single candidate has the opportunity to get interviewed. So over the course of a year, students speak with specology with with these avatars. Then they go and perform better on standardized tests.
>> What the banks are finding are that people are having an more comfortable time communicating about something personal like their finances with an AI human instead of a real human.
>> We'll have less stressed teachers, less stressed students, better grades.
>> I mean, there's nothing else like this right now. Using this this amazing technology, you have this one-on-one person you can meet in a safe environment. It doesn't feel like you're pressured. Similarly, [music] in healthcare, people are sometimes more honest with the AI [laughter] than they are with a real doctor that might impart some judgment. But I envision within 5 years that it will be as natural as going to the store and interacting with a [music] human. And we're very excited about what we can do together with H >> because we're helping education. We're helping society. And why create something if it doesn't help society? Woohoo. How's that? [laughter]
>> I I love seeing this. Every time it excites me. Okay, great. So, the last pillar um from me here is that I would think it's going to be a good idea to show you guys live of what it feels like to talk to a real avatar. Real avatar. [laughter] That's an interesting way to say it. Okay, cool. Between of >> Yeah. Okay. Great. Yeah. So, I'm going to try So, I'm going to be the homepage here. So, just to walk you guys through how to find the avatar to talk with. Um, so you can click get started for free. And once you log into this page, we'll see a lot of different avatars here and with different personas, different skin colors, roles, and we'll choose whoever would fit your persona and trying to have a conversation with them. Okay, let's talk with this lady. And it's called HR, meaning the knowledge base behind it's for recruiting. I'm just going to have a short conversation with her. Um, and so you guys have a feeling of what it feels like. Let's hope my connection is good right now.
>> Yeah, you might want to demonstrate uh live streaming on the live webinar. [laughter]
>> Yeah.
>> Okay, let's finger cross and see what it feels like. Okay, cool.
>> Hey, I'm Silus from Live Avatar. Nice to meet you. I'd love to chat about your background and what you're excited to work on these days.
>> Hi, Silus. How are you?
>> I'm doing well, thanks. How about you? How's your day going?
>> It's been fine. Um, I'm happy to walk through the agendas if you don't mind, and I'm happy to collaborate and answer your questions.
>> That sounds great. I appreciate your enthusiasm. Let's start with your current role or most recent project. What have you been working on?
>> Okay. I am a technical engineer on a very small platform team at a SAS company building an AI product.
>> That's interesting. What kind of engineering work do you enjoy most in that role?
>> Oh, I am designing and implementing backend services in Python and Typescript, maintaining our API layer currently.
>> Nice. It sounds like you have a solid foundation. Which technologies do you feel strongest in right now?
>> I think I'm pretty good managing cloud-based and databases, especially AWS.
>> Awesome. AWS is a powerful tool. Can you walk me through a recent technical challenge you faced while working with it?
>> Yeah, we do have an issue and bug in a production environment and I try to debug it. I look at logs, metrics, and recent deploys. I'll check dashboards for error spikes or latency changes and correlate it with code changes and try to reproduce the issue in a safer environment.
>> That's a solid approach. Debugging can be tricky. How do you usually optimize performance once you identify an issue?
>> Um, I usually will aim for a small, safe fix and follow up with a post-incident review to capture what happened.
>> That's a great strategy. It's important to learn from incidents. How do you like to collaborate with designers or PMs during projects?
>> Well, Silus, I actually I have something else that I got to go. Can I come back and reschedule with you later?
>> Of course. I appreciate you letting me know. Just reach out when you're ready to continue and we can pick up from where we left off. Have a great day.
>> You too. Bye.
Those are a very classic and standard u scenarios of LA avatar being an interviewer and then just to have a brief like information capture process um just give you guys a real feel like of what it it's really looks like. [laughter] So now I'm going to hand the mic to Chloe. Um, and so he's going to show you guys um how do how is our API and SDK look like. I'll stop sharing. Chloe, you can share your screen.
>> Yeah, thanks Sunny. I really tune in and really enjoy the conversation so far. Um, so can everyone see my screen? I think so. So like um Sunny mentioned, once you are in live avatar.com, how to get started is um navigate on these buttons and on this website. Before I dive into how you actually can implement with different um options like API, SDK, I want to go over functionality-wise. So here we have a list of 83 preset avatars that we actually hire professionals to film on specific sets of backgrounds. Hey, Chloe.
>> Yeah.
>> Sorry to sorry to jump in. I think we're still seeing the live avatar by Hen website instead of screen.
>> Let me share my entire window or >> Yeah.
>> Okay.
>> Thank you so much.
>> Yeah.
>> Always happens. Um, here you can see a list of public avatars you can leverage with given scenarios and given backgrounds that we have hired professionals to train and film on site. And those avatars are free for you to use for any live streaming sessions later on. Or you can has the option to create a custom avatar which requires uploading a two-minute footage of yourself or your spokesperson. And then Hun will then use that two-minute footage to train a digital twin of yourself that can be used later for dynamic conversations. And I also wanted to dive a little closer to what Sunny talked about on this avatar or any avatar that on the functionality-wise, we designed it in a way that it's highly configurable. So not only had does it has high-quality avatar presence, it you can also choose and configure the languages of your choice for this avatar to speak in. And similarly, you can also configure the knowledge base behind this avatar. So think of it as a contextual knowledge base behind the large language model um behind auto's text stack. And lastly, you can once you are configured with this um avatar, you can then there's multiple ways you can integrate with your applications or your website. First one is to um configure the avatar and then you can then copy-paste this front-end HTML element to your website. And so it behind the scene it will be an iframe um element that you can embed to your website. This is one of the lowest, I would say lowest development effort of implementations. Um, but if you want more customization capability and want more control, of course, we have integrates enterprise-grade API um list of APIs and SDKs available for you to integrate with. So um, if I navigate [snorts] here, or if you can go to docs.live.com, liveaver.com, you can see u the latest API docs that were written out for live avatar. If you have used hun before um the older version for this equivalent is streaming um streaming avatar. An overview of live [snorts] avatar is basically regardless of API integrations or SDK integrations, we're actually trying to provide and serve different personas or different uh use cases depending on the customer's technical resources. So we we of course offer a low-code approach that I shared with you on better versions. We also have a pro-code or high-code approach where you can integrate with our API or SDK. And we designed the API in a way that it's similar to UI that is highly configurable. You can choose avatar. You can configure which avatar looks or appearance um like um for that visual layers. And for the voice and audio layer, you can define the voice quality, the voice motion, the speech, um the pitch or the um even more like higher customizations around voice settings that we cover with our voice bender. And of course, like um like any large language model where allows you to define how your avatar thinks and responds to control, for example, the personal person uh personality traits. For example, like we showed you um, you can attach a knowledge base around hiring, recruiting, or component. You can also attach um specific product knowledge for growth training. Uh, either way, it's highly configurable and it can be passed as just an API parameter. And lastly, I want to talk about um live avatar session, which is the fundamental core of how we orchestrate the entire conversation. Um, so with that overview, I want to touch point like at a high level, there are two ways we two modes we provide for users to integrate with regardless of using SDK or or using API. The first one is called full mode and the second one is called custom mode. At a high level, um, I will I will rate full mode as more um less code needed or u less orchestration needed for among different vendors versus custom mode will allow you to uh choose your own vendor for speech to text, for large language model, for voice activity detection, and for um text to speech t uh stack versus the full, you kind of um delegate this entire text stack orchestration to Hunen and let us handle the heavyweight heavyweight lifted parts for you. And you handle um the the configuration of the avatar, the knowledge base, and the voice to make it super natural to interact with users. Um, at the core of each conversation, it's the live avatar session lifecycle. I won't go much into this detail to bore you right now. Um, all the details are here available in the docs.live.com for you to read. And to get started, you can actually go to our quick start session. And it only takes three steps basically um to first have your API key. Once you log into live avatar and make sure you are on at least free tier above to have a valid session to start. Once you have the API key, you will then create a session token. This is a a short-lived um one-time access token. We want to make sure that every conversation, every session is uh it's private and it's protected by a short-lived access token following the standard of um design in the industry. And then you can then um command to start the sessions given the session token that you got. And the third step would be um joining a LiveKit room. LiveKit is actually um one of our vendors using to uh orchestrate the live streaming sessions over the internet. And we found that LiveKit is actually way has better latency and performance on um real-time conversations over the internet than other vendors. To get started, you can then you can go to our GitHub. Um, we uh actually sorry, let me show this instead. Um, so to get started, I think this will be very straightforward for any developer to just clone this project, this um web SDK. And then if you can if you wanted to see how it looks like immediately and host on your localhost, you can go to apps-demo and then following the readme to install all the required packages and just run um the sample application. Looks like um, I wanted to see show a real example. Okay, I think this is actually that's that's it. Sunny, I wanted to share how edu, our live customer, actually implemented this, but looks like the website right now it's not reachable.
>> Oh, is it?
>> Yeah.
>> Okay. There's the problem. Yeah, it works for me though. Anyways, we can cover the Q&A right now, Audrey, if we want.
>> Yeah. You guys want to do some We've got lots of questions if we want to jump into the Q&A.
>> Yeah. Is it on a community page?
>> Yes, we've got a list of them. I think Sunny, I see you in the list, too. Um,
>> Yeah. Okay. So, we just answered the questions there, right?
>> Okay. Wonderful. Um, so Chloe, I'm gonna I'm gonna answer the questions that I can answer here and you probably would chime in.
>> Um, so I got some questions. The first is, will latency ever get to sub 100 ms? Um, in the very near weeks, no. [laughter] So um, we're shooting for um the the the best latency like for for sure. And we can from the uh uh to the avatar layer, we can control it below 500 ms. But for 100 ms like as a perceivable latency, that's not doable for now. But as the technology evolves and I think we're we're definitely going to see better latency. And we're trying really hard to make we're leading this in leading the industry in regard of that. Okay. So the second is, I want to connect the live avatar with my learning management system. Uh, ideally the avatar can interact with students within the learning environment. Is it possible to build this? Yes, that's exactly what it is for. [laughter] So it can after you connect to the avatar with your um system, it can definitely interact with your students. And many of our customers already have done that. So um, I think next is your question, Chloe.
>> Yep.
>> You want to read it through?
>> Yeah.
>> Yeah. The next question is uh recent service outage. What happened? Customer avatars would fix the creation process. Data privacy with large language model TTS support for dialects could tune live avatar map. Um, recent service outage, the one day I'm aware, I think we reported at our status uh system status dashboard. Um, the outage happens can vary depending on what factors. Um, because um live avatar streaming [snorts] is such a technically challenged technology stack that we're reliance on multiple vendors behind the scene on top of our own premises model. And so for example um if one of the vendors is down, then we could be impacted as well. We're trying to resolve it as quickly as possible. Um, so I don't know this specific outage that this user mentioned about. For data privacy with large language model, so we are we are all compliance with security compliances like GDPR um and other compliance rel uh related to the US. And you you can find out at security.hen.com um data privacy regarding large language model. I can I can share that the large language model behind the scene, it's open to models um the default model it's 4.0 or mini. So those models as you can already know, like these as long as they're compliant with data privacy regulations, and we will also comply as well. Support of dialects is similar [clears throat] to functional um the multilingual support that we just showed you. So yes, we support that. Kun live avatar, Sunny, I don't think >> Um, we can well, so let me handle this. So we can do it. Um, but we do not recommend um it's it's like based because Hagen's technology and including our avatar model is trained based on humans' face and so it's like a lot of dimensions on a human face, right? So the facial expressions, the very natural movement are there. So but cartoon or any like AI-generated avatar won't be able to have that fluency. Um, well, we can do it, we can make it work, but it's going to be the quality is going to be a little bit polar than the real person's one. So you feel free to try, but we do not recommend that. It's that is to say um it's not something that in a very close roadmap either. Um, but yeah, it it can work though. Yeah. Okay. Um, I think the next is question is Chloe as well.
>> Yeah. What's the most reliable way of connecting avatar to external knowledge base um using you're using Python and Chroma database for vector embeddings? So I would say in this scenario, you will want to go to custom mode where you choose your own large language model and you define your external knowledge base there. Um, and then you will then connect to your large language model to the live streaming sessions by passing us the your large language model response. And we we can coordinate and stream either that's an audio response or a text response. So, Hun was still handling the visual rendering layer for the avatar to speak the response coming out from your large language model, regardless of what text stack you use. Um, next one. Will people be able to import existing HAN avatar to live avatar at some point in the future? Question about migration. Um,
>> Um, I think it's mostly about um the video avatar to streaming avatar. Um, so I got this question a lot um so far, no. And because video avatar at hijen and live avatar are two different products. It's it's so different from the product layer and the infra layer. So we haven't enabled that um but do we want to? Of course. And so it's on the roadmap um but currently no. So you have to build your own avatar in the live avatar website separately. By the way, the website is livear.com. Make sure you guys check it out. Okay, cool. Next is questions. Can live avatar conduct quiz and tally response say four questions with each having A B C D? Can it be integrated with Typeform or Tally? Well, it can do some tasks, but I'm not sure if it can integrate with Typeform. Um, so, uh, have the avatar to do tasks is something that we put on a very close roadmap. It probably going to be ready in one or two months. It's um, the avatar should be able to pull out transcript and pull out documents for you after a simple conversation. And if you wanted to build a a question or questionnaire, and you can set in your knowledge base, but not necessarily in the form of Typeform, but probably it can still conduct the questionnaire. Um, so the next question is, I'm an agency. What is the best way to build and manage this for clients? Um, well, so it's nothing different than like building it for yourself. You have to connect to the API yourself. I think what is the most important thing is to have a brief idea of your usage, like how many clients do you have and how many streaming minutes of each of them would request for the coming years. And have that usage is very important because that impacts the the price, right? So it's very uh intuitive that the higher you consume um the lower the price would be. So have the best estimation of projection of the usage would be the only uh differentiate like oh sorry. The only factors that for your scenario, otherwise it's all the same, right?
>> And I would say also depending on your client's use cases and their functionality requirements um our API SDK is highly customizable depending on how you want to orchestrate all the API sequence. So definitely try out the API demo. Yeah. Okay. Next question is, it seems like previous custom streaming avatars do not work with live avatar. Is that correct or should I be able to use it? Um, so assuming the avatar you built in the interactive avatar space, right? So, we will have a migration process in the next few weeks that we will send out to all the uh interactive avatar users to help them migrate to live avatar. So far, it's still not connected because we just launched. Just bear with us for the migration plan. We promise this is going to be delightful and smooth. Okay, cool. And I next question, I teach com I teach communication and I think this is misspelled something between words and body language is very important, like hand gestures. Is that problem solved? And can you edit video to correct it? Um, so the true answer is no. [laughter] Hand gesture is on the roadmap. Um, but it's not something we prioritize right now because most of the Hijun avatar and we focus on chest up. The movement is very natural out there. We can do full body, we can do basic hand gestures, but you can't say, "Okay, I want to be a poker uh starter." or you want to have a very sophisticated hand gesture. That's not doable right now. So, I'll say we focus on the conversational layers as of now. Yeah. How can I make my live avatar less expressive? And is there a sheet of common demands so I know the correct format to use? You wanted your avatar to be less expressive. [laughter] That's that's rare. Okay. So, um, you can put that request in your prompt and say you want the avatar to be cold, >> to be more >> Yeah. to be more shy or something. And you can actually architect the persona a little bit. Most of the LLM should be able to apply that prompt to to to shape the reaction. So, yeah, the short answer is you can. And how does this compare to something like Tavis? Okay, Tavis is our dear friend in this space. Um, so the short answer is but better. [laughter and gasps] Competitor but >> yeah. Okay. So next question is, before a few months ago, I was on an older pricing plan and now I wonder if maybe allow both API to custom avatar in one in one place or in one price um or what changed with the plan changing? >> They want to consolidate the pricing. >> Okay. Still separate. So build a custom avatar is a separate effort then enable the streaming. So streaming would still be streaming and avatar would still be avatar. So far. Um, it's nothing changed. The logic has been the same. All right. All of the examples I've seen so far are user-driven. The user asks a question and controls the interaction. I want the avatar to teach a course. So the avatar is in control. The course the avatar teaches is in the knowledge base, but it can be interrupted to answer questions. Possible? Yes. [laughter] Um, so as long as you upload your knowledge base and kind of prompt it the way that you want the avatar to talk through it, and yeah, the students can pause anytime and the avatar should be able to diagnose then when and who is talking to it. It will pause and answer the question. Okay. I tested interactive avatar during beta. Uh, latency, ST, LLM, TTS limitations made serious usage impossible. Are use cases like AI recruitment truly production-ready or still mostly polished demos? Well, it was like two years ago and now it's definitely production-ready as I showed you. And the use case that we worked with Zlegate. Um, and they have already used it for and then 14,000 interviews have been already conducted. And we have saved people 24 hours per hire. And I think that's a very strong and powerful proof of that it's we can actually provide serious commercial value. All right, next. Live avatar via hijen API to my applications. Is hunen ingesting any of our proprietary data into its foundational LLM? You want to cover that, Chloe?
>> Yeah. So, if I connect live aar via hun API to my applications. Um, so yes. So, remember we shared two modes of integrations, full mode versus custom mode. On the custom mode, where you will be relying on Hun's large language model vendor. And of course, your data um included in the conversation will be transmitted along the way to hijen. But if you are concerned about your proprietary data, which I understand it's likely the case for uh financial services industry or healthcare industry, you can always choose to integrate with choose your own large language model and have your external use u knowledge base configured there. And then uh agent will not um be able to to get those sensitive data. So custom knowledge base is the way to go for this scenario. Next one. Can live avatar work with OpenAI realtime API? Um, I actually don't know the answer or maybe it's not documented. But live avatar essentially, it's a better version of our original streaming avatar. And with the streaming avatar API, we do have the [snorts] option to integrate with OpenAI real-time API for real-time audio streaming. Um, yeah, can check us out the docs.
>> Okay, I'm going to say god the the question last so long. Thank you for the enthusiasm, guys. Um, so does um so assume that I am using the API but I want to use my own avatar. What do I need to purchase? It's a great question. So um, you need to purchase your own custom avatar, which is uh $50 per month. Um, and you were like after connecting the API, like you have an estimate of how many streaming minutes that you need and purchase the streaming minutes separately. All right. It's all available in self-serve. So you can check it out livear.com. Okay. And does the avatar actually pick up from where you left during your next call since its memory is not refreshed for the new call? That's a nice catch. Yeah, that was just the words and I feel it's a nice way to hang up. Um, so we don't carry memory as of now, like you said. So next time if you wanted to conduct a new interview, you have to start from scratch. So make sure you leave enough time for that interview to be finished. Um, because I cut it off because I don't want to take too much time demoing on the call. Um, that was just a you know, nothing for uh real memory. Yeah. Is memory, Sunny? Is memory on the roadmap? I think there was another question. Yes.
>> Yes. Yes, it is. But yeah, I'm gonna I want to be transparent about what we're capable right now. So we don't set the expectation too high, but it's on the roadmap for sure.
>> Awesome.
>> Yeah. Does the avatar act also I have a scenarios use the avatar inside of a web page embed into authoring tool? Is that possible? I think so. Right Chloe? Yeah. Yeah.
>> Yeah. This is possible for sure. And will the interactive avatar APIs and functionality ever be discontinued or deprecated in favor of live avatar? Um, so far it will still be on, but um we will gradually sunset it. So please after we send out those migration instructions, please uh kindly migrate everything to live avatar because we are doing all the updates and in live avatar going forward. So you should be have the access to the most like the newest features there. Um, but interactive API would still be on for a while. Uh, but we do recommend to migrate. Okay. And can two avatars interact with one another like home shopping network or QVC type of style? H so far we don't prioritize that use case. Uh, we're still focused on one-on-one or one-to-many like avatar conversations. Um, we don't have anything established for two avatars interacted with each other. But if you feel like the your use case is very significant and the usage is already there, and we we can explore or provide some guidance to you in terms of how to build it. As far as I know, this is not a technology that's feasible easily now um but but we can talk about it if you feel it's important. Um, can two avatar inter Oh, so pricing still is confusing with Hen have multiple pricing pages?
>> I think it's time to share the pricing page uh directly on the web app.
>> Um, yeah, but it's I'm I can talk through it because it's quite simple, right?
>> Um, if you are building avatar for clients, how to monitor token use per project? Perhaps put limits. Okay. Yeah, I think so too. So, Hijen does have a lot of different pricing layers because we have different products, right? And different products serve different logic. So that's why we have different pricing. And we wanted to make sure we provide the best price value to our customers. Um, so bear with us. Um, so the thing is for live avatar self-serve um we have different modes like like um Chloe mentioned. Oh, let me show you my page actually. That would be easier, right? Uh, there we go. So, going back to the homepage and once you clicked um get started, and there is a little present here, pricing. You click here. This is our different tiers of pricing and free 10 credits to start. And this is in central tier with all the functions that we can support. And go down here. This is SDK's price. So have full mode and custom mode. Full mode means that you have everything managed by hijen >> and you just use it and it's two credits per minute. But for some of our customers because they already have the vendor relationships with other tech stack, for example, the TTS, LLM, ASR, and they wanted to keep it and we also wanted to provide that flexibility for them and they only wanted to use the as a avatar interface. It's absolutely possible. So that's custom. So and for self-serve, it's 0.1 per minute. Um, so and a custom avatar, like I said, is $50 per month. And for enterprise solutions, um, it's custom based on your usage. So feel free to contact sales here um at Hen and we can talk about it. All right, great. Okay, next questions. We wanted to cover all of them before we wrap up.
>> I think we're we're 10 minutes over time, but yeah, we can cover up. Um
>> Um, Audrey's time allows us to cover all of them.
>> I don't know if we'll be able to get through all of them. Let's run. Do you guys both have another five minutes?
>> Yeah. Yeah.
>> Okay. Let's see.
How many we can get through in five minutes. I think we've gone through a ton of the questions. Um if you are in the chat and your question didn't get answered, um let us know. We'll look through those. Um and we'll share resources after this as well.
>> Cool. I'll try my best. Uh we could do memory our our end with the custom option, right?
>> Yeah.
>> I think so.
>> The car large language model and custom knowledge base.
>> Yeah. And what is the time estimate for when the migration guide will be available?
In the next few weeks, [laughter] I promise you. Yeah. I I we I got question a lot and we're really heads working on it because we wanted to make sure it's simple enough to do.
Okay. Um I have one avatar and I use it for learning topic. Can I use also for topic B for both group A and B can still use it? I think this is really depends on your knowledge base. Okay. Is knowledge base universally true. If it's true then you can do it but if you like you want a physical teacher to teach English I would recommend you not to do it. Okay. To build another avatar maybe it would be easier.
>> Yeah. And building knowledge base is free. Doesn't consume credits.
>> Yeah. Is a new live avatar model going to have the same business model as the former interactive avatar?
What does that mean? So I kind of didn't catch that question. Um
>> the question is um is the new live avatar model going to have the similar cost and limitation than the previous versions?
limitation in terms of why
>> um yeah well we might need some clarity there [laughter]
>> but overall I would say live avatar it's a better versions of interactive avatar in many um cases uh we make the cost a lot simpler and then depending on whether you are coming from small startup like um with selfert plan or enterprise for custom high volume usage base um limitation ations in terms of scalability concurrency we have significantly improved um than the previous interactive avatar infrastructures and I think education plan it's more of like um like a startup plan like
>> what is education plan
>> maybe with discount I I would assume so [laughter]
>> try are essential
>> yeah I would say always start with our self-s serve plan.
>> Yeah.
>> Yeah.
>> Is there any way to get a stored version of the conversation audio? So all the transcript will be in text format
>> like we don't offer audio format. Um so it should be sufficient for any analytics or you wanted the information captured but we don't have audio.
um have built a four-step verification HR recruitment workflow. Can I integrate this to my automation workflow? If yes, do advice on process? Um I'm not sure what you mean by recruitment workflow. Um does that mean that you want to check some box or Yeah, I I'm not really sure about that. uh maybe need more clarification on that question but uh mo most of the times if you if you have a knowledge base and you have the logic of how the avatar should pick up the question it's totally doable
>> can live avatar be deployed in the mash network to edge devices
>> very tricky so um I will say though the the technology of live streaming is very heavily depending on the network the um the end clients like your browser, your PC and depending on what region you are starting from um starting the streaming sessions. So I would say you will um experience some latency issue if you are doing it on the mesh network and over an edge um what is it edge devices? [laughter] Yeah,
>> I saw one question from the YouTube. Um, so I think I wanted to pick it up. Say, I'm one of your first customers love what you were doing. I just wish you guys have a more accessible pricing plan. It's still very prohibitive for small companies. I wanted to touch base on the pricing thing a little bit. Um, because as everybody know that live avatar like streaming anything streaming avatar is very expensive, right? It's very complicated on the technology. Um what we're doing right now is that we're trying everything we can to um reduce the cost and so to reduce the price. So the price will change as the technology evolves. Um but just please trust us like we're doing everything we can to make sure the price is the best. Uh but we are not advocate ourself as the cheapest plan and that's not something we're proud of. We are shooting for the best price value for our customers. Um so price-wise we're trying um but it's not something um universally cheap um technology
>> depending on what your comparator if you're compared to other products like video translations from Hen streaming is actually a lot cheaper. Um but keep in mind that this is very cutting edge technology and we're doing a lot of heavy lifting and orchestration among different models for you and it's happening all instantly over internet. So the the cost of course we want to make it as cheap as possible but we also cares about the quality and the performance on latency. Um yeah
>> yeah overall you will get cheaper and cheaper over time.
>> Yeah. Well, please check out liveavvatar.com, okay? And see how you can connect it with your own business.
>> Awesome. I think that's a good place to end. Um, thank you for going through all of those questions, Sunny and Chloe. I think we got through a ton of them. Um, a lot of thank yous in the chat for the session. Um, I know we always there's so many amazing questions. um we get through as many as we can and then there's lots of resources in the chat where you can post additional questions, you can find self-s serve resources um and more to come from this team. I think we'll be back on for another session in the future. So looking forward to that and hope everyone has a great day.
>> Thank you guys.