Riva Enterprise#

From version 2.0 onwards, Riva is free of charge for both development and production use up to a daily usage limit. For enterprises looking to deploy at a scale above the daily usage limit for ASR and TTS, we announced NVIDIA Riva Enterprise; a paid offering that includes enterprise support.

Riva Enterprise includes:

  • Annual usage license of Riva ASR and TTS on all cloud and on-prem platforms

  • Access to NVIDIA AI experts with enterprise business support for deployment, configuration, and performance questions

  • Support for all recent releases to control your maintenance and upgrade schedule

  • Priority access to new languages, software patches, upgrades, and maintenance releases

World Class Speech AI with Assurance of Enterprise Grade Support

NVIDIA Riva Enterprise Short-Term Supported Releases#

The Short-Term Support (STS) offering of Riva Enterprise provides AI innovations for teams building services and products leveraging Riva Enterprise. Service providers, ISVs, and other organizations can leverage this program to access NVIDIA support for releases in the last six months starting with NVIDIA Riva 2.0. Refer to the Release Notes for releases within the last six months.


To enable the Riva Enterprise metrics collection when deploying a Riva Speech AI container, set the following environment variables:

  • RIVA_API_KEY set to the NGC API key associated with your account

  • RIVA_API_NGC_ORG set to the organization or organization/team within NGC to associate metrics with. The organization must be one you are a member of.

  • RIVA_EULA set to accept to acknowledge that you have read and agree to the terms specified in the Riva End-User License Agreement.


The Riva server will not start if the API key is specified but the EULA is not accepted or an organization is not supplied.

Quick Start#

The Quick Start deployment path can be used for testing Riva Enterprise. Refer to the Docker deployment configuration section for details on the configuration options available. Within config.sh, uncomment the Enterprise section and fill in the variables appropriately:

# Enable Riva Enterprise
# If enrolled in Enterprise, enable Riva Enterprise by setting configuration
# here. You must explicitly acknowledge you have read and agree to the EULA.
RIVA_API_KEY=<ngc api key>
RIVA_API_NGC_ORG=<ngc organization>


The Helm deployment path can also be used for testing Riva Enterprise. Refer to the Kubernetes deployment section for details on the configuration options available. Within values.yaml, enable the Enterprise section and fill in the variables appropriately:

  ngcSecret: rivadeploysecret
  ngcOrg: your_org_name
  acceptEula: "accept"
  enabled: true

Frequently Asked Questions#

Q: For Riva ASR, how is usage measured?

A: Riva ASR usage is based on the amount of speech audio successfully processed (not including audio that does not contain speech). The Riva VAD algorithm described in Beginning/End of Utterance Detection is used to detect periods of speech vs nonspeech.

Q: For Riva TTS, how is usage measured?

A: Riva TTS usage is based on the number of characters sent to the service to be synthesized.

Q: What kinds of data does the NVIDIA Riva server Docker images send back to NVIDIA?

A: The following data is reported to NVIDIA.

Personally Identifiable Information (PII) captured data includes:

  • Customer Universally Unique IDentifier (UUID)

  • Customer organization and team

  • GPU

  • Container UUID

Customer telemetry captured data includes:

  • Audio and speech duration

  • Request count

  • Total characters

  • Release version