Sonic от Cartesia — это сверхбыстрая ИИ-модель для синтеза речи с рекордно низкой задержкой всего в 135 мс. Она создает живые и естественные голоса в реальном времени, что делает её идеальным выбором для интерактивных приложений и голосовых ассистентов.
We've used this architecture to train a new voice model called Sonic that we're releasing today. Sonic creates high quality lifelike speech for any voice with a model latency of 135ms—the fastest for a model of this class. We've built and optimized our own state space model inference stack to enable us to serve Sonic with low latency and high throughput—enabling us to serve high quality models at lower costs. Sonic is released with a web playground and low latency API. The playground features a diverse voice library for applications across customer support, entertainment, and content creation with support for instant cloning and voice design (speed, emotion), all of which can be used through the API.