I was surprised this summer when Alexandru Voica, head of corporate affairs at video generation startup Synthesia, sent me a link to a new member of his PR team. It’s his interactive virtual avatar, trained to answer frequently asked press questions about Synthesia, such as what it does and how it works. The day before, I was on a panel and a publicist asked me if I would mind pitches using AI-generated text. But Alexander’s avatar went beyond that. To me, this seemed like the final boss of using AI in PR.
In September, Synthesia invited me to visit their new office space in New York. Originally based in the UK, Synthesia is an up-and-coming digital avatar startup alongside D-ID, HeyGen, Colossyan, and more. The company reached a valuation of $4 billion earlier this year and announced ARR of over $100 million last year.
Synthesia allows companies to create interactive training videos using AI avatars, and recently launched a product called Role-Play Sessions that allows employees to practice sales pitches and more using interactive AI avatars that respond and score.
When I went to the opening of my new office, I was asked, “Do you want your own AI avatar?” Without hesitation, I said, “Yes.” Of course I want my own digital twin. My outfit that day was cute and my hair was coiffed.
I had no interest in avatars until I discovered digital twins, but I felt they would inevitably become a part of my online life. I’ve heard that some people on Instagram are creating portraits of themselves to help create social content. I find all of this very interesting, and maybe that’s why I have no hesitation in sharing my digital twin with you now. This is the first time Synethsia has created a digital avatar for a journalist (or anyone other than Voica). This article was trained on my story about why venture-backed startups have more fraud than non-VC-backed startups, and will only answer questions about that story.
Below, you can simply press “Start in new window” to get started.
You can ask questions such as:
Why did you decide to write this story? What is the research paper about? What did the researchers discover?
To make this, I went into a mini film studio inside Synthesia’s offices, where I took a bunch of photos of me and recorded two minutes of my voice. You had to consent to the creation of these avatars, and the Digital Dom was born. They created a personal avatar for me (one that just reads the script I gave them) with and without glasses, and two interactive avatars (one that can return conversations and listen to me) with and without glasses.
We selected articles to train an interactive avatar, and one of the teams built my interactive avatar using a combination of speech-to-text, video, language, and text-to-speech models. My Avatar’s tech stack includes Synthesia’s proprietary video and audio models, but the company also allows customers to choose alternative models from other labs such as Cartesia, Celebrities, Google, and OpenAI. Businesses can also choose to host their avatars on their preferred cloud or pay Synthesia to do it for them.
Speech-to-text models convert what people say into text, and agent language models can understand the text and take actions based on it. A text-to-speech model converts the responses to audio, and finally a video model (built by Synthesia) animates the avatar’s conversations.
Overall, Synthesia is building three types of products. One is a video creation and distribution platform with a classic avatar, where someone enters a script and the avatar repeats it. An agent platform called Sessions that allows people to interact with avatars through surveys and role-plays. and an API platform that allows people to take Synthesia’s video and audio models and combine them with other technology services to build interactive avatars and other types of products.
It took the Synthesia team several days to create my avatar. I played with something personal first and typed in a fairly generic script to see how the AI’s voice sounded. I asked them to talk about my favorite time of the year, autumn, which has arrived in New York. The audio was fairly accurate and I was pleased to see that it didn’t pick up any of the hoarseness that occurred when recording audio samples.
I showed it to a non-technical friend and he found it both interesting and creepy.
Next, I showed them the interactive robot that I had created. In this case, that was my venture fraud story. Every time I asked questions like where I was before TechCrunch or what part of New York I lived, I was directed back to the story.
My friends didn’t think the voice sounded very similar to mine, and didn’t think it was as good as my personal avatar, but it was still close enough to feel a little creepy. My mother also praised it, calling it “amazing.” She and my father kept asking “only they know” questions about me, but the model wouldn’t answer and redirected them to the venture scam story each time.
After testing the model, she joked, “I don’t remember having you two.”
This experience made me think about what the future of journalism would be. Do people agree that when they turn on the news, their avatars will show them the news?One investor quickly said no. Indeed, there is a lot of backlash today against the development of AI that has permeated social media and other news-sharing platforms. But others I asked weren’t so sure. Can avatars augment or replace journalists? Would a CEO want to talk to an AI avatar of a journalist instead of a human avatar?
My favorite things about my career are connecting with people, writing stories, and researching new topics. But the biggest part of journalism is trust. It’s unlikely you can outsource that to AI.
Even outside of journalism, I’m sure the idea of cloning yourself is appealing. No more catching up on work after vacation or vacation. Because a version of yourself will always be there to answer your questions.
We’ll have to wait and see how the use of avatars evolves in corporate America. But now that I have my own, I have mixed feelings. After I got over the initial novelty, I waited for the Avatar to stop talking, then I studied it carefully. I don’t know what it was for. Maybe it’s waiting for it to flash. Or it could say something new, smile, or just let me know that it knows.
My digital twin is deterministic, so that will never happen. However, with an AI powered by a chatbot that is non-deterministic, meaning it can be overbearing at will, we find it easy for someone to slip into something like AI psychosis.
I told an investor that I predicted that whatever the future holds for digital twins, my generation, Gen Z, will never get used to them. They feel sci-fi. Everything you’ve ever seen in movies is here now. However, I think digital avatars are less offensive than humanoids. At least with an avatar, you can always log off if things get weird.
But until then, here’s my personal avatar, bringing you all the top stories on our site this week.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
