Nvidia Riva

• About NVIDIA® Riva is a GPU-accelerated SDK for building Speech AI applications that are customized for your use case and deliver real-time performance.

• Models -

• Languages - Riva supports English (en-US, en-GB), Spanish (es-US), German (de-DE), Russian (ru-RU), Hindi (hi-IN), French (fr-FR), and Mandarin (zh-CN). here.

Our connector allows you to leverage the power of Nvidia's Riva SDK using our engine. With the help of Riva, you can run production-grade conversational AI inference for tasks such as speech recognition, speech synthesis, and a variety of natural language processing inferences.

Streaming Speech Recognition

• Streaming
• Offline / Batch
• Word-level timestamps
• Top-N transcripts (Alternatives) from beam decoder
• Multiple speech recognition models deployed simultaneously

Diarization

• Streaming (Coming Soon)
• Offline / Batch (Coming Soon)

NLP

• Text Classification
• named entity recognition
• Joint Intent + Slots
• Extractive Q&A
• Punctuation and Capitalization

Riva offers several methods to hone transcript quality. Using this connector, our engine takes care of leveraging your data to carry-forward context and learn from past data, to obtain context and domain relevant transcripts, that improve in quality over time and usage.

Transcript Refinement

• Word boosting
• Custom vocabulary
• Language Model retraining
• Acoustic Model fine-tuning