Computer vision applications are getting a big boost from advancements in smartphones combined with AI that helps fill in the gaps that everyday devices can’t see. A Munich startup called Beyond Presence believes that using these tools holds the key to our future. In other words, they are hyper-realistic avatars that look and sound exactly like humans and can be used in real-time conversation situations.
The startup announced today that it has raised its first external funding, a $3.1 million pre-seed round from multiple backers. German company HV Capital is leading the round, along with 10x Founders, Alba VC and individuals from Meta, DeepMind and Zalando.
Beyond Presence hasn’t actually released its technology into the world yet, but it’s coming soon, Awais Shafique, the startup’s CEO and co-founder, told TechCrunch.
He said some of the new capital will be used to continue developing the company’s basic model, and some will fund a public beta launch by the end of the month.
So far, about 300 companies are on the product waiting list, with early applications likely to apply in areas such as customer service and support, recruiting, sales, and e-learning. These are all areas where companies want to expand their interactions with users without having to hire and train more people.
The founder’s background serves to explain how the startup raised millions of dollars in funding before even launching a product, and how the resource-intensive AI company got to this point without any external funding.
Shafique previously co-founded Presize, a computer vision startup that can accurately measure a user’s measurements by recording a clip of them glancing back in front of their smartphone camera. These measurements can be used to purchase clothing online. The Munich-based startup gained huge recognition in its home country because its founders participated in the German TV version of “Dragons Den”/”Shark Tank” in 2020, raising a record €650,000.
Then, in April 2022, it was revealed that Meta had acquired Presize. The transaction price and exact date were not disclosed, but some signs point to a big deal.
Subsequent SEC filings showed that Meta spent between $774 million (January through March) and $1.15 billion (through the first six months of the year) on acquisitions. Presize was the only deal made public at the time. Meanwhile, Wikipedia’s list of Meta acquisitions has been compiled with a much more modest estimate of $100 million.
Sources familiar with the matter told TechCrunch that in this case, a deal value in the low nine figures would be more accurate than a 10-figure figure.
Regardless of where Presize falls on that spectrum, considering the startup has only raised “$2 million,” by Shafique’s estimate, he and his co-founders have had a very successful exit.
“It’s profitable enough that we don’t have to work,” he said in an interview.
(They’re all still working on it anyway. In addition to Shafique’s latest startup, Presize’s other co-founders, Leon Szeli and Tomislav Tomov, are still working on their generative AI project at Meta. But they’re still in touch with Shafique and invested in Beyond Presence. I am doing it.)
At its core, Presize aims to solve challenging computer vision problems, and Beyond Presence aims to do just that.
Of course, my initial thoughts on Beyond Presence’s proposal were skeptical.
Digital avatars aren’t exactly uncharted territory. Not only have they been around for decades, but they have recently been in the spotlight with the craze for all things chatbots, with the latest innovations in AI and processing.
Nor do we have a clear idea of what will work in the avatar space as a long-term business. Some, like Hyper and Ready Player Me (supported by Amazon and A16Z, respectively), build on gaming culture and the theory of what an avatar should be. ~ no Be a realistic version of the people who use it. People who want to hide their identity to protect their privacy may choose this approach over replicating an exact likeness.
Other companies like Synthesia ~can do Create figures of people. However, this requires a visit to a studio and is not the basis of the main product, which uses a simple device like a phone camera to capture facial expressions to power the avatar. ~ no It looks like you.
And some people are trying to build tools that allow us to interact using everyday devices. These include Veed and Deepbrain.
Beyond Presence aims not only for avatars that can be created on the fly using your phone, but also for avatars that resemble you and can be used in real-time, unscripted conversation situations. It’s like a video version of text-based chatbots from OpenAI, Anthropic, Meta, and others.
Of course, startups are not the only ones in this field. Zoom appears to be developing hyper-realistic custom avatars, and CommonGround is also developing something along similar lines.
Currently, Beyond Presence is focused on building something unique in the video component of its product.
Eleven Labs is supporting the voice part of the experience, and OpenAI’s GPT is a generative AI model supporting the content. Business customers have the flexibility to change the model they use.
“You can also leave it all to us or use the LLM of your choice or any voice agent of your choice. “Then I draw a face on it,” he said. For example, this feature addresses the fact that corporate clients are already working with a specific LLM provider or do not want to share their proprietary data externally.
There are some interesting details about Beyond Presence’s initial decisions regarding funding. Although the company was accepted to participate in Y Combinator earlier this year, it has since withdrawn from the renowned incubator.
YC certainly provides access to a great network, but Shafique and his co-founder Felix Altenberger (CTO, former deep learning expert at Presize) are determined not to give up equity and autonomy to access it.
YC Partners may have had their own opinions about how to grow the company through sales, but Shafique said there was a “horizontal” approach he and Altenberger wanted to pursue.
“We wanted to move more from the core technology and foundational model layers,” he said.
