Speech and Translation AI
Build and deploy fully customizable multilingual speech and translation AI for your large language model and retrieval-augmented generation based applications.
NVIDIA® Riva is a set of GPU-accelerated multilingual speech and translation microservices for building fully customizable, real-time conversational AI pipelines. Riva includes automatic speech recognition (ASR), text-to-speech (TTS), and neural machine translation (NMT) and is deployable in all clouds, in data centers, at the edge, and on embedded devices. With Riva, organizations can add speech and translation interfaces with large language models (LLMs) and retrieval-augmented generation (RAG) to transform chatbots into engaging, expressive multilingual assistants and avatars.
Achieve high multilingual transcription and translation accuracy, and provide out-of-the-box, expressive, professional female and male voices with state-of-the-art models pretrained on thousands of hours of audio on NVIDIA supercomputers.
Customize across ASR pipelines for different languages, accents, domains, vocabulary, and context for the best possible accuracy for your use case and across TTS pipelines for the brand voice and intonation you want.
Provide consistent experiences to hundreds of thousands of concurrent users with higher inference performance than existing technology, and deploy anywhere—in data centers, on premises, in the cloud, at the edge, or in embedded devices.
Accelerate the development and deployment of production-grade, multilingual, voice-enabled AI applications with NVIDIA AI Enterprise, an end-to-end, cloud-native software platform for enterprise-grade secure and stable generative AI.
Experience new ASR, TTS and NMT microservices now available—designed to provide optimized AI inference for speech and translation AI. This includes Parakeet models that deliver recording setting ASR accuracy and performance.
See how NVIDIA AI supports industry use cases, and jump-start your speech AI development with curated examples.
Companies are deploying AI virtual assistants to automatically address the queries of millions of customers and employees around the clock. With Riva’s speech and translation AI microservices, these assistants provide helpful and natural responses at every turn of the conversation despite background noise, poor sound quality, and diverse speaker dialects and accents.
Consumers expect contact center agents to resolve their issues quickly and efficiently. To meet these expectations and deliver the best customer and agent experiences possible, enterprises across industries are implementing agent-assist technology powered by Riva speech and translation AI.
To enhance customer service experiences and build strong relationships with their customers, businesses are building avatars with recognizable brand voices. With Riva, they can create a unique, high-quality, personalized voice with just three seconds of speech data.
With hundreds of millions of online meetings held daily, video conferencing has become an indispensable tool for enterprises. Through Riva's real-time transcription, video conferencing applications achieve impressive accuracy in live captioning and meeting summarizations, accommodating users with worldwide accents and diverse domain-specific vocabularies.
In the global economy, businesses operate across countries and serve customers with diverse linguistic and cultural backgrounds. This diversity in global languages poses a unique challenge, as hiring native speakers and training employees in multiple languages isn't scalable, cost-effective, or efficient. Riva translation empowers accurate and effective communication, facilitating smooth global interactions.
AI robots are increasingly found in hospitals, airports, and retail stores worldwide. They aid frontline workers by handling daily repetitive tasks in restaurants and manufacturing facilities, assist customers in locating items in stores, and support physicians and nurses in patient care. With Riva, it’s easy to add speech and translation AI to service robots.
Use the right tools and technologies to build and deploy fully customizable, multilingual speech and translation AI applications.
Use the right tools and technologies to build and deploy fully customizable, multilingual, speech and translation AI applications.
Explore everything you need to start developing with NVIDIA Riva, including the latest documentation, tutorials, technical blogs, and more.
Talk to an NVIDIA product specialist about moving from pilot to production with the security, API stability, and support of NVIDIA AI Enterprise.
In 2021, AI2Labs spun off from Yoozoo Games as a local tech startup in Singapore. AI2Labs innovates, experiments, and develops AI products and applications, enabling efficient processes and improving sustainability and business outcomes.
AI2Labs integrated Riva into their Speakr—domain-specific speech AI—speech recognition API to accommodate the intricacies of Asian speech and business domains and achieved state-of-the-art Singlish translation accuracy.
Avaya specializes in cloud communications and workstream collaboration solutions, providing unified communications, contact center, communications platform as a service (CPaaS), and services with their OneCloud platform.
Avaya integrated the NVIDIA Riva speech-to-text engine for real-time captions at scale. Riva enables better transcription quality, lower word-error rate, and economical delivery.
For over 10 years, the Applied AI Group at C-DAC in Pune, India, has focused on research and development of speech technology. They’ve successfully created a cutting-edge speech-to-text (STT) system for Indic languages such as Hindiand Marathi. The group continues to advance their work by exploring AI-enabled, open-source deep learning frameworks, libraries, and tools for creating STT and speech-enabled applications for other Indic and low-resource languages. Experiments were conducted using various neural network architectures and topologies from NVIDIA’s open-source NeMo framework, with Citrinet and Conformer-CTC network topologies proving to be effective in building and training neural acoustic models for speech recognition. These models were trained on single- and multi-node Param Siddhi AI systems, optimizing training time and performance. Finally, the models were deployed for real-time and batch-mode inference using the Riva GPU-accelerated production pipeline.
NCS, a subsidiary of Singtel Group, is a leading technology services firm with presence in Asia Pacific and partners with governments and enterprises to advance communities through technology. Combining the experience and expertise of its 12,000-strong team across 61 specialisations, NCS provides differentiated and end-to-end technology services to clients with its NEXT capabilities in digital, data, cloud and platforms, as well as core offerings in application, infrastructure, engineering and cybersecurity. NCS also believes in building a strong partner ecosystem with leading technology players, research institutions and start-ups to support open innovation and co-creation.
NCS uses NVIDIA Riva TTS in Breeze—the driver’s companion app—for voice-guided navigation, live traffic and road condition updates, real-time parking rates, and electronic road pricing rates and operating hours, to help Singapore drivers experience smooth driving journeys.
RingCentral, a leading provider of global enterprise cloud communications, collaboration, and contact center solutions, serves millions of users. The RingCentral platform empowers collaboration from any location and device, improving business efficiency and customer satisfaction. RingCentral uses NVIDIA Riva for video conferencing transcription for 200,000 concurrent users on their platform.
Snap is a camera and social media company that enables multimedia message creation with filters and effects. To create more interactive experiences, Snapchat users play with Lenses—a feature that adds real-time effects into snaps—over 6 billion times per day.
NVIDIA Riva’s noise- and lingo-optimized speech AI service is integrated into Snap AR Lens Studio, enabling creators—artists and developers—to build gripping augmented reality (AR) experiences.
T-Mobile, a supercharged Un-carrier, delivers an advanced 4G LTE and transformative 5G network for the best customer experience. To empower contact center agents, T-Mobile implements Expert Assist. This AI-based software uses NVIDIA Riva to transcribe real-time customer conversations that feed recommenders and assist thousands of agents.
With Riva, T-Mobile fine-tunes automatic speech recognition models on custom datasets and interprets customer jargon accurately across noisy environments.
Use the right tools and technologies to build and deploy fully customizable, multilingual, speech and translation AI applications.
Explore everything you need to start developing with NVIDIA Riva, including the latest documentation, tutorials, technical blogs, and more.
Talk to an NVIDIA product specialist about moving from pilot to production with the security, API stability, and support of NVIDIA AI Enterprise.