Case Study: Inworld achieves 70% faster time to first audio with Modular

We made state-of-the-art speech synthesis scalable, and achieved a truly remarkable improvement both for the latency and throughput

The faced significant challenges in creating a scalable, low-latency, and cost-effective speech synthesis service. Their text-to-speech with Inworld to co-engineer a solution using its MAX Framework and the Mojo programming language. This integration optimized the entire inference stack, resulting in a state-of-the-art speech pipeline. The results to first audio, achieving 200ms for the first audio chunk, and a 60% lower cost for their API, enabling Inworld to serve

Inworld

Igor Poletaev

Chief Science Officer