AWS enables custom development of Alexa voices

Get a free Techzine subscription!

AWS has brought the Brand Voice feature to its Amazon Polly service, enabling companies to develop custom voices for Alexa Voice Assistant.

The AWS Polly department is engaged in Artificial Intelligence (AI) research. Among other things, neural text-to-speech (NTTS) technology is being developed to make artificial voices sound like human beings. Brand Voice uses deep learning models to discover intonation patterns in natural speech, after which it reproduces the voice. The AI model is able to do this by consisting of two components. A number of results can already be heard in the announcement.

Results

AWS now helps companies to develop custom voices using Brand Voice. National Bank Australia and fast food chain KFC are among the first customers to test the service. At KFC, for example, Colonel Sanders’ voice is being replicated. The deceased iconic founder of the fast food chain will be used to respond to customer queries. In a statement, the KFC states that “the Colonel” was passionate about his fried chicken, and that Brand Voice enables ordering in a “fun and memorable way”. National Bank Australia takes Brand Voice very seriously, and the organisation sees it as an opportunity to improve customer experience.

AWS has previously released similar features for Alexa devices. For example, it was able to recreate the voice of Samuel L. Jackson. The actor didn’t have to have his voice recorded for hours for that. The NTTS models were able to convert audio recordings of Samuel L. Jackson into a re-production of his voice.

Other efforts by Amazon also show that it wants to make Alexa more and more a technology for business application. For example, Alexa for Business Blueprints was previously unveiled, which should make it easy for employees to create new skills for Alexa.