Paris-based voice AI company Gradium has launched Voice Design, a feature that generates new synthetic voices from written descriptions without requiring reference audio. The tool is now live in the Gradium API and Studio across five languages.
- Users input 1 to 500 character descriptions detailing attributes like gender, age, accent, and pitch to receive up to 5 voice candidates in 3 to 5 seconds.
- Candidates are generated via a four-call REST flow, where selected voices can be promoted to permanent custom voice slots using one slot shared with clones.
- In blind pairwise listening tests across 7,627 comparisons, Gradium achieved a 72.6% win rate against competitors like ElevenLabs and Inworld.
- The system is non-deterministic, meaning the same prompt yields different results each time, and unsaved candidates expire after 30 days.
This allows developers to create specific voice personas on demand without sourcing rights or consent for real speakers.