WellSaid Launches Natural Synthetic Voice Tech for Creators
WellSaid Labs emerges from the AI2 incubator with a new voice synthesis system that embraces human inconsistencies to create natural-sounding speech. The startup aims to provide a credible alternative to human voice actors for creators, bypassing the restrictive tech offered by major tech giants.
WellSaid Labs introduces a new approach to synthetic speech that aims to give creators access to natural-sounding voice technology. Unlike major tech companies that reserve their advanced voice models for virtual assistants, WellSaid focuses on making high-quality synthetic voices available for applications like screen readers and automatically generated audiobooks.
The startup emerges as the first graduate of the Allen Institute for AI incubator program and builds upon existing neural network research like Tacotron. Instead of relying on heavy human annotation to force a perfectly consistent and robotic sound, the founders intentionally design their system to embrace the natural inconsistencies found in human speech.
To achieve this authentic sound, the team collaborates with professional voice actors to record dozens of hours of audio for the system to learn from. By prioritizing these organic variations over computationally cheap, rigid models, WellSaid successfully creates a highly credible synthetic alternative to real human voices.