How ReadSpeaker Synthesized Breaking Bad’s Giancarlo Esposito’s Voice for Sonos’ New Voice Assistant
The new Sonos voice assistant speaks in the voice of actor Giancarlo Esposito, known for his roles on Breaking Bad and The Mandalorian, using ReadSpeaker’s synthetic voice technology. The digital voice interface creator unveiled the role of the ReadSpeaker VoiceLab in designing the new Sonos Voice Control (SVC), the latest high-profile brand turning to ReadSpeaker for a bespoke custom voice AI.
“We started the project with Sonos about a year ago. They audited the voice technology market and reached out to us because they understood that our VoiceLab unit could really accompany them throughout the custom voice creation,” ReadSpeaker CMO Roy Lindemann told Voicebot in an interview. “Either a brand comes to us and wants us to take over and find the voice talent that matches what the brand wants or they already have an idea in mind. In this case, Sonos came up with Giancarlo Esposito and asked us to work with him to create a digital carbon copy of his voice that could be leveraged in the voice assistant.”
Esposito signed on with Sonos to serve as the voice of its AI, then worked closely with ReadSpeaker to record enough of his voice to train the Deep Neural Networking (DNN) models used by ReadSpeaker. The models could then say words and phrases Esposito never recorded. He recorded specific scripts and coordinated with computational linguists to ensure his speech style stayed consistent regardless of the words. The non-standard text and proper names necessary for reading out track and artist titles made that aspect especially crucial, according to Lindemann.
The voice assistant includes both English and Spanish-speaking modes, further necessitating precision. Though not in the immediate pipeline, Esposito’s voice may eventually converse in several more languages. ReadSpeaker’s catalog includes several dozen languages already, most recently Ukrainian, so it would be within the startup’s wheelhouse to further enhance Esposito’s artificial linguistic skills.
“We selected ReadSpeaker to power the custom voice for Sonos Voice Control for a variety of reasons, not least of which was their ability to work closely with us from the development phase all the way through delivery,” Sonos vice president of voice experience Joseph Dureau said. “ReadSpeaker’s approach to custom voice is innovative and extremely thorough – which is exactly what we required for this project – and the company’s dedicated linguistic team has been a critical tool for us throughout this project.”
ReadSpeaker has been rapidly accumulating new features and partners over the last year. The Sonos partnership mirrors the revelation that ReadSpeaker handles Spotify’s interactive voice for its Car Thing automotive audio player. The company then made a move to the metaverse by adding a real-time text-to-speech plugin for both the Unreal and Unity game engines in February. Game developers can embed a single AI-generated audio setup in a game running on multiple platforms. That followed ReadSpeaker’s unique voice user interface for the Nintendo Switch video game system. That same synthetic voice modeling system is even training pilots to fly the F-35 Lightning II fighter jet.
“We pitch ourselves as a digital design company and design is the important word there,” Lindemann said. “Big companies expect a company like us to really identify the voice talent that is coherent with their brand positioning. Our expertise is also important as not every voice talent will convert to good synthetic speech. They need a team to assess the pool of voice talent.”