Voices Launches Character Audio Dataset for Voice AI Training

Voices, a voice solutions and voice data provider, has released a character audio dataset featuring character-specific voice data for voice artificial intelligence training.

The dataset offers a range of voice data, featuring 19 tones, 35 emotions, and 21 paralinguistic cues across more than 450 character types, ranging from a 1920s Flapper, a Drill Sergeant, a Wise Old Wizard, and more.

Voices character datasets are delivered in a ready-to-use format. Each audio file is accompanied by rich metadata tagged by character, text, tone, and emotion and fully transcribed. All audio recordings were collected using Voices' virtual Recording Studio. The recording process includes auto-directed sessions.

"Voice data is both scarce and in high demand. The base models powering voice AI today simply cannot deliver authentic sounding, tailored experiences. The best voice AI models will be built on curated and ethically sourced character data, which is exactly what we have launched today," said Dheeraj Jalali, chief technology officer of Voices, in a statement. "Voices has a long legacy of working with professional voice talent. Given that history, we are uniquely positioned to connect trained contributors with emerging AI applications, and as the industry grows, we see even more opportunities for voice talent to help meet the market demand."