About the Project Project Hydrus is a scripted and conversational audio data collection project. The recordings produced help build natural, everyday-sounding synthesized voices for conversational AI products — voices that sound like real people talking, not
Direct voice performers through scripted scenarios, coaching for natural, conversational delivery rather than verbatim recitation. Build quick scene context for performers (who, where, why) and improvise to fill narrative or dialogue gaps live, keeping sessions flowing