Advances in smartphones combined with AI are giving huge boosts to computer vision applications, filling in the gaps of what everyday devices cannot see. Using those tools, a startup out of Munich believes it holds the keys to what is coming next: Hyper-realistic avatars that sound like, look like, and communicate like their human counterparts and can be used in real-time, conversational situations.
The startup today said that it has secured its first external funding: a pre-seed round of $3.1 million. German firm HV Capital is the lead investor in the round, together with 10x Founders, Alba VC, and a few individuals from Meta, DeepMind, and Zalando, among others.
Beyond Presence has not released its technology into the wild yet, but it's close, said Awais Shafique, the startup's CEO and co-founder, when speaking with TechCrunch.
This money, part of the new capital, is planned to be used to continue the development of the company's foundational models, and some will be in funding the public beta it will be releasing by the end of this month.
So far, about 300 companies are on a waitlist to get the product, which will likely see early application areas in customer support and service, recruitment, sales, and e-learning. All these are areas where companies want to scale the interaction with the user without necessarily hiring and training more people.
So how, exactly does a startup raise a multimillion-dollar round even before it has launched its product and come this far as a resource-intensive AI company without taking any external funding? The answer is in the backgrounds of the founders.
Shafique had previously co-founded Presize, a computer vision startup that can take a good measurement of a user by filming a clip of someone spinning around in front of a smartphone camera. Those measurements can then be applied to guide the same one in buying apparels online. The Munich-based startup had a relatively oversized profile at home due to the fact that its founders appeared on the German version of "Dragons Den"/"Shark Tank" in 2020, raising a record-breaking €650,000.
And it was then in April 2022 that reports came out saying Meta acquired Presize. Neither price nor date of the transaction was revealed, but there are some good reasons that show how well this exit would be.
By SEC filings, between January and March 2022, Meta spent between $774 million and $1.15 billion in the first six months of 2022 on acquisitions. That was when only one deal, Presize, had been publicly disclosed. Wikipedia's list of Meta acquisitions, however, has been edited with much more modest estimates: $100 million.
Sources close to the deal said that the figure for this was in the lower nine figures, not 10 figures as some reports had put it. Whatever the place of Presize in that range, with the company having only "raised a couple of million," according to Shafique, he and his fellow cofounders did pretty well to exit in the first place. "It's lucrative enough that we don't have to work," Shafique said in an interview.
(They're all still working anyway. Beyond Presence's other co-founders, Leon Szeli and Tomislav Tomov, both stay at Meta and develop generative AI on its behalf. And they keep in touch with Shafique, because both of them are actually investing in Beyond Presence.
At its core, Presize was trying to solve a troublesome computer vision problem, and that is what Beyond Presence is trying to do, too.
I must confess that, once I first found out about Beyond Presence, the pitch didn't really seem something I could buy into.
Digital avatars, after all, aren't an unexplored territory: they have been around for decades and more recently gaining attention, maybe even hype, in recent innovations of AI and processing-not to mention the craze for all things chatbot.
Nor are there clear notions of what will be successful in the world of avatars as a sustainable commercial endeavor. Some, such as Hyper and Ready Player Me, which is backed respectively by Amazon and a16z, are tapping into gaming culture and the notion that avatars should not be realistic portrayals of the people using them. Users who wish to obscure their identities for the purpose of maintaining privacy may prefer these models over those that resemble them exactly.
Others, like Synthesia, can create images of people, among other technology that also uses ordinary devices like phone cameras to take your expressions to animate avatars that do not look like you. For now, at least, though, these cannot be used to facilitate real-time, live interactions.
Others that are also building tools that you can engage with using ordinary devices include Veed and Deepbrain.
Beyond Presence has even more ambitious goals-that the on-the-fly creation of an avatar using your phone also looks like you and is used in real-time, unscripted conversational situations. Comparable to a video version of those text-based chatbots offered by OpenAI, Anthropic, Meta, and many others.
Of course, this space is not dominated by one startup. Zoom, for instance, seems to be working on hyper-realistic custom avatars, and CommonGround is also developing something along those lines.
For now, Beyond Presence focuses on building something unique in the video component of its product.
The voice part is powered by Eleven Labs, while OpenAI's GPT is powering the generative AI model for content. Business customers are able to switch between which model they want to use.
You can just let us do everything, or use any LLM that you want, or any voice agent that you want. And then we put the face on top of it, he said. That feature addresses the fact that enterprise customers may already be working with particular LLM providers, for example, or do not want to share any proprietary data externally, he noted.
Beyond Presence' Early Funding Decisions: In an interesting twist, the company was accepted into Y Combinator earlier this year but declined the spot in that prestigious incubator.
While YC undoubtedly offers some fantastic networks, Shafique and his co-founder Felix Altenberger (CTO; and a former deep learning specialist at Presize) chose not to sacrifice equity and autonomy in service of access to that.
While YC partners might have been able to insist on a method through which the company could scale by sales, Shafique said he and Altenberger had a "horizontal" approach they wanted to pursue.
"We wanted to move more on the core technology and foundation model layer," he said.