Voicemod has created 20 “AI humans,” digital personas that allow gamers to speak in voices other than their own.
Voicemod
An artificial intelligence firm in Spain has unveiled a number of “AI humans” that allow people to modify their voices in video games — in real time.
Voicemod, which is based in Valencia, on Wednesday released 20 humanlike characters, ranging from a 20-something woman to an elderly man. The voices are trained on voices from professional human voice actors.
In a demonstration of the tech on a call with CNBC, Voicemod’s CEO, Jaime Bosch, showed himself speaking regularly and changing it to several different voices, from a high-pitched female speaker to a baritone male voice.
Gamers can download the app on their Apple Mac or Windows PC and incorporate Voicemod’s technology as a “virtual microphone” that sits in between the microphone application they’re using to start speaking through the alternative personas.
Voicemod, which counts talent from leading technical universities in Valencia and Barcelona, has been working on voice synthesis and interactive audio features since 2014, with many of its employees specializing in music technology and audio.
Used by more than 40 million people, Voicemod’s tech can be used by people in the social app Discord to speak with each other in voices other than their own while playing games.
“We have an amazing creativity community using this mainly to have fun, enjoy it with their friends and have a sense of belonging,” Bosch said in an interview with CNBC.
“One of the biggest use cases I love the most is shy people — we have some people who write to us saying I wasn’t able to really socialize with people who are now able to do that.”
It’s a milestone in the world of conversational AI. While many of today’s AI algorithms allow people to submit text and receive something AI generated back, the feat of ensuring this is done in real time is much harder.
The technology requires a significant amount of computing power. And producing and patenting the proprietary algorithm models behind it takes a lot of investment, and talent.
To that end, Voicemod has raised $23 million in cash from several venture capitalists including Leadwind and Bitkraft Ventures.
Bosch is also no stranger to the risks surrounding how the technology can end up being abused — voice-changing technology could be used to imitate leading political figures or scam people, for example.
“This is something I think of every single day, something we’re thinking about in the company every single day when it comes to the creation of voices,” he said.
The company is close to finishing a “watermarking” solution that can identify whether voices have been generated using modification systems. It is also in discussions with other firms about standardization of such systems and ensuring voice-changing tech is released safely.
“One thing is, clear laws will come,” Bosch said. “We know Europe is working on that. The reality is that, usually, companies go faster than the laws.”
WATCH: Cannes Lions 2023: AI won’t make markets lazy — TBWA CEO
