Aura offers around a dozen voice models, all trained in-house with a dataset created with voice actors. The pricing for Aura is competitive, beating most of its competitors at $0.015 per 1,000 characters. The speed and quality of Aura's speech-to-text model are also notable. Deepgram has been building the underlying infrastructure for four years before releasing the product, focusing on achieving a good price point, low latencies, and high accuracy.
Key takeaways:
- Deepgram has launched Aura, a real-time text-to-speech API that combines realistic voice models with a low-latency API, aimed at helping developers build real-time, conversational AI agents.
- The AI agents, backed by large language models, can be used as customer service agents in call centers and other customer-facing situations.
- Aura's pricing is competitive, beating most of its competitors at $0.015 per 1,000 characters, which is slightly cheaper than Google’s WaveNet voices and Amazon’s Polly’s Neural voices.
- Aura offers around a dozen voice models, all trained in-house by Deepgram using a dataset created with voice actors, and its speed and accuracy are highlighted as standout features.