Aura-2 adds Dutch, French, German, Italian and Japanese
The text-to-speech model now supports five additional languages for enterprise use.
Aura-2, a text-to-speech model, now supports Dutch, French, German, Italian and Japanese.
Deepgram said in a statement these languages join existing English and Spanish models to extend the company’s multilingual voice infrastructure and to provide consistent experiences via its API.
According to Deepgram, each language presents distinct phonological and prosodic challenges: Dutch requires handling vowel richness and long compound words; French needs liaison and elision management; German demands clear enunciation of consonant clusters; Italian preserves open vowels and musical intonation; and Japanese requires pitch accent handling and seamless script switching.
The company said Aura-2 is engineered for business use, offering sub-200ms latency, high-precision pronunciation for structured data such as IDs, times and currency, reliability under streaming loads, and a simplified infrastructure that removes the need to combine vendors. Developers can switch to any new language by updating the language code in their API request, Deepgram added.
Deepgram pointed users to its TTS Voices and Languages documentation and the Deepgram Playground for audio samples and testing, and said new sign-ups receive $200 in credits, enough to generate 13 million characters of synthesis. "The goal is clear: voice AI that works everywhere, for everyone," Deepgram said.
The Recap
- Aura-2 now supports five additional languages for TTS.
- Adds Dutch, French, German, Italian and Japanese language models.
- Developers can change language via the API language code.
