Resemble AI today unveiled Rapid Voice Cloning, an exciting new feature of its platform designed to rapidly create voice clones. Resemble AI specialises in artificial intelligence voice solutions specifically tailored for enterprise users.
Rapid Voice Cloning, now available, enables users to duplicate voices from relatively short datasets in just about one minute – marking an important development and making voice cloning technology more accessible, empowering more users with custom voices for their applications and having an impactful effect across content creation, personalization and accessibility. According to Resemble this development will make an immediate and lasting impression across fields like content creation, personalization and accessibility.
Resemble published several voice cloning samples to demonstrate its capabilities, while VentureBeat conducted tests of it as well.
How Does Resemble’s New AI Voice Cloning Feature Work? Resemble offers users the ability to clone their voice digitally using its web platform by uploading an audio sample or recording multiple sentences. Although Resemble had offered this feature before, creating one took time – users typically needed to record 25 sentences or upload at least three minutes of content before setting up their system, with final cloning taking up to an additional hour or more after that initial process had completed.
Now with Rapid Voice Cloning’s release, users can more quickly get started with this technology. All they need to do is provide an audio sample ranging from 10 seconds to one minute of their target voice; then the company’s model captures all parameters including accents instantly from this sample and gives back results within one minute for downstream use cases.
Resemble AI’s advanced machine learning algorithms excel in replicating the subtleties and accent nuances, according to its blog post on Rapid Voice Cloning. By learning from just 10-second voice samples, Rapid Voice Cloning creates AI-generated voices which perfectly replicates an original speaker’s unique intonations, pronunciations and cadences of their accent.
The company published numerous samples comparing its offering with Microsoft’s VALL-E and XTTS-v2 voice cloning models, complete with input voice sample and text used for cloning. Results were impressive. But when we created a free test account to see how the tech actually worked for ourselves, there were noticeable gaps.
In our tests, Rapid Voice Cloning required recording three long sentences without offering an option to capture shorter 10-second samples. Processing was quick but could not recognize our Indian speaker’s accent properly – taking instead what sounded like American English as input instead. Unfortunately this affected the output voice’s accent; though Rapid Voice Cloning claims support for most English accents.
