Amazon is the most recent tech large to unveil a voice AI mannequin. In accordance with Amazon, its Nova Sonic is “a brand new basis mannequin that unifies speech understanding and speech era right into a single mannequin, to allow extra human-like voice conversations in AI purposes.” Nova Sonic will compete with comparable AI fashions by OpenAI, Google, and different tech corporations.
Nova Sonic understands greater than phrases
The Nova Sonic doesn’t simply perceive the speaker’s phrases, however it could actually additionally course of the tone, type, and tempo. The AI voice generator adapts to the dialog context, so dialogue flows extra naturally, in comparison with the extra stilted fashions from the primary generations of Alexa. The Nova Sonic can do that as a result of it combines a number of speech processing and producing capabilities right into a single AI mannequin as a substitute of utilizing a number of completely different fashions.
Historically, AI voice instruments concerned working a number of fashions in sequence: a speech recognition mannequin would convert speech to textual content, then a big language mannequin (LLM) would course of the enter textual content and generate responses, and eventually a text-to-speech mannequin would convert textual content again to audio. This complicated pipeline usually stripped away the tone, type, and pacing of the speaker’s unique dialogue.
Because the Nova Sonic combines all of this in a single mannequin, it could actually adapt to the acoustic context of the enter speech. It additionally responds extra naturally to the cadences of human speech; for example, it received’t interrupt when the speaker hesitates or pauses to take a breath.
The best way to get Nova Sonic
Nova Sonic is at present accessible by way of a brand new API in Amazon Bedrock, the corporate’s enterprise software constructing platform, and can simplify the event of voice purposes.
What builders must learn about Amazon Nova
The tech large not too long ago launched Amazon Nova Act, a brand new AI mannequin skilled to carry out actions inside an online browser. As well as, there’s an Amazon Nova SDK for builders to discover. One of many basis fashions is Nova Canvas for producing high-quality photos; there are additionally fashions for producing textual content from completely different modalities in addition to movies from textual content and picture enter.
