Interview with CEO Barnier Geerling

Interview with CEO Barnier Geerling

Interview with CEO Barnier Geerling





Dutch startup DAISYS develops the AI-driven voice technology of the future.

With its entirely in-house developed AI technology, DAISYS can 'compose' lifelike human voices down to the smallest details. This gives the Dutch startup a crucial lead in the fast-growing international billion-dollar market for chatbots, interactive games and voice-controlled cars, TVs and other consumer electronics. "We are developing the Holy Grail of AI-driven voice technology here," says DAISYS CEO and co-founder Barnier Geerling.

"The human voice is the interface of the future ", says CEO Barnier Geerling of Dutch AI startup DAISYS."Voice control is increasingly taking over the role of buttons, keyboards and touchscreens. Whether you are producing an interactive video or game, a personal AI assistant or a new car, TV or coffee machine: soon you won't be able to do without a voice interface. That voice has to be able to communicate with the user in just the right tone, timbre, emotion and rhythm."

Unique technology

So the total user experience then hinges on the right voice. "That has to represent your brand, service or product perfectly," says Geerling. "It has to be pleasant and clear, invite interaction, and preferably completely tunable to the user's preferences and real-time emotions. The ability to compose voices entirely according to one's own preferences is therefore very valuable.And that is precisely the unique technology we have developed over the past four years."

Rapid development

As a voice actor, voice director, sound studio owner and "real science fiction nerd", Geerling has followed the rapid development of voice technology closely for years. "From a tinny robot voice, thanks to the rise of machine learning and AI, it suddenly went at lightning speed to real human-sounding voice. That was the moment I jumped on that with my two partners.But with our own, and completely unique, approach."

Cloned LEGO cubes

The technology of the moment uses databases of existing voices. Algorithms distort or 'clone' that voice to create the illusion of a new voice. "But so that 'new voice' is really just a modified version of an existing voice," Geerling observes."As a voice actor, I know better than anyone how sensitive that can be.So I wanted to go a step further, and develop technology that can completely artificially compose every aspect of the voice."

Technological challenge

Recent court cases, such as that of Scarlet Johansson against OpenAI, currently demonstrate the rightness of Geerling's feel-good preference. At the time, however, Geerling is saddling his partners and rapidly growing team with a huge technological challenge. "The complete unbundling, or disentanglement, of source data and output is something of a Holy Grail in AI land," he laughs. "At the time, I absolutely did not oversee how complex that would become.But we did succeed in the end."

Unique distinctive value

The unique, AI-driven voice technology that DAISYS now presents differs from the services of any other provider in a number of crucial ways. First, the startup compiled its own data set of human voices - indispensable starting point for any AI voice technology - entirely by itself. Geerling: "Sixty Dutch voice actor colleagues lent their voices, and received a small share in return. In this way, we show directly that we take responsible technology very seriously."

Melody bows

DAISYS also developed the required machine learning and AI technology entirely in-house. "That can now completely isolate every component that makes a voice unique," says Geerling. "So not just timbre or rhythm, for example, but also features for which there was simply no name until now. We therefore now invent terms like 'melody bending' ourselves. Users can therefore compose their desired voice themselves with extreme precision. That is a real game changer."

Automated learning process

Users will soon be able to compose their unique voice (or have it composed) via DAISYS' SPEAK platform, also developed entirely in-house. Moreover, with a smart API access, they will connect the platform directly to their own back-end technology. "In the future, chatbots will, among other things, be able to 'measure' the emotion of the interlocutor," Geerling explains. "Via our API, the chatbot can then automatically compose a voice that responds in a soothing or happy tone, for example."

New data platform

That capability greatly increases the quality and effectiveness of AI-driven communication. "Now that this unique technology really stands, we are going to further train our model with large acquired and open source data sets," Geerling continues. "To do that quickly and effectively, however, we need to 'pre-process' that data. Fortunately, we have now also developed a platform that can completely automate this laborious task. So we are now really ready for the future."

Reputable venture capitalists

DAISYS has already worked on pilots with major Dutch broadcasters, and is in talks with a major Dutch insurer. The startup is now also talking to investors to fund further development of the technology, the now 15-strong team and the available data set. UK market leader ElevenLabs, which works with the less advanced cloning technology, raised USD 80 million from several reputable venture capitalists in January this year.

Sober Dutch approach

"Thanks to our down-to-earth Dutch approach, we need much less money," laughs Geerling. "But it is crucial that our technology now quickly learns a lot about all the different accents, dialects and other challenges that characterise the English language worldwide. We have a big head start at the moment, but given the lightning-fast developments, it could quickly shrink. Fortunately, our unique approach does enable very efficient training of our AI models."

Rosy future

So now a Dutch startup is producing the Holy Grail of AI voice technology. That a golden future lies ahead for DAISYS, which recently moved to larger premises in Amsterdam, thus seems very plausible. "The voice market within our reach accounts for over $37 billion," says Geerling. "If we can capture just a small part of that, we are already on a roll. Thanks to the power of our unique technology, we really dare to dream of that now."

Discover more articles..