Currently, voice services exist in various scenarios, including real-time meeting records, real-time live video captions, call center voice quality inspection, and real-time meeting records. This year, NVIDIA released NVIDIA Riva, a ready-made voice service that can be easily deployed in any cloud or data center. The service can process hundreds to thousands of audio streams as input and return text with minimal delay; it can also quickly build high-level conversational AI services. NVIDIA Riva is an SDK that uses GPU acceleration to quickly deploy high-performance conversational AI services and can be used to quickly develop voice AI applications. The Riva SDK runs on NVIDIA GPUs and provides the fastest inference response at high throughput levels. Currently, NVIDIA Riva integrates intelligent algorithm engines such as ASR and TTS, and users can use these functions for scientific research.