Over the past several months, Liopa has been working on a major technology breakthrough that creates audio of people’s speech from silent videos. Where lip movements are visible in a video, speech can be generated with Liopa’s technology. The breakthrough is based upon our Visual Voice Generator solution.
The AI-based solution will read a speaker’s lips, re-creating their speech using a computerised voice – or their own voice.
“In terms of commercial availability – this is a world first,” said Liam McQuillan, Liopa’s CEO. “There are so many routes this could take us down – and the important thing is that we’re the first company able to launch an AI-based speech creation technology.”
The following two-minute video demonstrates the technology:
Liopa’s CTO and co-founder, Fabian Campbell-West explained further: “People can send us any video of someone speaking – as long as their lips are visible in the video – and we can return a version with the speech recreated as a computerised voice. Or, they can send us an existing sample of the person’s actual voice, and the system will tailor the output to be similar to that voice.”
There are many instances where people take recordings without realising that the audio is not being recorded. It’s a common mistake with interviews and business meetings – even informal family videos. This happens over cloud solutions like Zoom, but also on desktop applications such as Apple Keynote.
“With many recording technologies, you need to hit a separate button saying ‘Mic on’. People don’t realise this, and they can record an entire interview, conversation, or something important, and realise at the end they taped a silent video,” said Campbell-West. “This happens so frequently that the market is crying out for a solution – and we’re hoping to jump on this opportunity.”
Liopa has been inundated with requests from people searching for a solution that can re-create lost sound. The technology breakthrough will fulfill a demonstrable market need.
“Furthermore, any video could be used – whether it was taken in 2023 or 1943,” said McQuillan. “There are endless applications for this technology going into the historical record, as well as for modern-day needs.”
People who have silent videos who want to be an early tester should send a query to firstname.lastname@example.org. Please look out for our official website and product launch in early 2024.