Gradium

Gradium is the voice layer built for real-time applications, with models and voices optimized for conversational AI and voice agents. Its models cover text-to-speech (TTS), speech-to-text (STT), voice design, voice cloning, and live translation.

TTS combines 50 ms ultra-low latency with 0 ms of leading silence by default to support natural conversational turn-taking, without drift or audio distortion as traffic scales.

Designed for reliable pronunciation of transactional data, it handles phone numbers, email addresses, dates, times, currencies, web addresses, and alphanumeric confirmation codes. Language-aware rewrite rules and custom pronunciation dictionaries give teams control over brand names, SKUs, and specialist terms.

The catalog features over 1,400 voices, with consistent performance across supported languages and regional accents, including American, British and Irish English, Québécois French, Bavarian and Austrian German, Castilian and Mexican Spanish, and Brazilian and European Portuguese.

Gradium Voice Design creates a distinctive vocal persona from a text prompt in a few seconds, Voice Cloning reproduces an existing voice, and localization adapts these voices to other languages and regional accents for tailored voice experiences in different markets.

Gradium is ISO 27001:2022 certified and SOC 2 Type II attested, with EU and US data residency, Zero Data Retention available, and flexible deployment options.