Spotify is aiming to add emotion to your listening experience through a new system that synthesizes text-to-speech. The technology takes written text and converts it into audio that captures the emotion and intention of the original text.
The system works by feeding text into an AI synthesizer, which then converts the text into speech data. This data is then processed by a neural network-based Vocoder, adding in emotional cues and accents indicated in the original text. The result is speech that sounds more natural and human-like, capable of conveying emotions like anger, happiness, or sadness and accents like French or British.
This technology has been trained on datasets of audio samples and corresponding text until the output sound is realistic enough. Spotify has previously shown interest in AI voice technology, having acquired Sonantic, an AI voice platform, last year.
The technology has potential beyond delivering music. For instance, it could generate emotionally engaging audiobooks. Spotify launched its audiobook offering in the US last year, now boasting more than 300,000 titles.
However, there are concerns about the technology’s potential misuse. Any AI voice simulator that can produce a realistic voice is potentially vulnerable to being used for deep fakes. Furthermore, it could impact voice actors whose livelihoods depend on their unique voices. Despite these potential drawbacks, the innovation is seen as a significant step forward in enhancing the user experience.