NVIDIA® Riva is a GPU-accelerated software development kit (SDK) for building and deploying fully customizable, real-time multilingual speech and translation AI pipelines including automatic speech recognition (ASR), neural machine translation (NMT), and text-to-speech (TTS). Riva delivers world-class out-of-the-box accurate transcriptions and translations, and engaging professional voices. It can be deployed on-prem, in all clouds, embedded, and at the edge.

With NVIDIA Riva customers can customize for different languages, accents, domains, vocabulary, and context to achieve the best possible accuracy for their use case, and the brand voice and intonation they want. It provides consistent experiences for hundreds of thousands of input streams with higher inference performance versus existing technology.

To accelerate building speech-AI-based AI solutions, Riva includes packaged AI workflows for audio transcription and intelligent virtual assistants that include pretrained models and resources such as Helm Charts, Jupyter Notebooks, and documentation. Riva is part of the NVIDIA AI Enterprise software platform.

NVIDIA Riva AMI includes:

• NVIDIA AI Enterprise Catalog access script
• Access to Riva container on NGC
• Ubuntu Server 22.04
• NVIDIA vGPU Driver
• Docker-ce
• NVIDIA Container Toolkit
• AWS CLI, NGC CLI
• Miniconda, JupyterLab (within conda base env), Git