Gradium TTS (Sep 2026)

Gradium TTS is a real-time text-to-speech model developed by Gradium, designed for low-latency voice applications and interactive AI agents. Released in September 2026, the model minimizes audio generation delay while improving expressive prosody and native handling of complex text structures.

Core Capabilities

The model features ultra-low latency, achieving fast times to first audio, and includes robust built-in handling for difficult tokens such as phone numbers, email addresses, dates, and alphanumeric codes without requiring separate text normalization. It supports multi-lingual synthesis across English, French, Spanish, German, and Portuguese, alongside instant voice cloning capabilities.

Performance and Integration

Optimized for streaming cloud use cases, Gradium TTS provides high-precision word-level timestamps to ensure synchronization between text and generated audio streams. Benchmarks highlight its high accuracy on pronunciation hard-cases, making it applicable for responsive conversational voice assistants.