MODELS

Deepgram Nova-3

by Deepgram

modelsource:deepgram.comspeech-to-texttranscriptionstreamingbatchmultilingualDeepgram

Overview

High-accuracy Deepgram speech-to-text model for batch and streaming transcription, with multilingual support and self-serve terminology customization.

Details

Deepgram Nova-3 is a speech-to-text model documented by Deepgram for high-accuracy transcription. The supplied Deepgram sources describe it as supporting both batch and streaming transcription, multilingual use cases, and self-serve terminology customization, making it relevant for applications that need configurable speech recognition across recorded audio or live audio streams.

When to Use

Use Nova-3 when you need a Deepgram speech-to-text model for high-accuracy transcription of recorded audio or batch jobs. Use Nova-3 when your application needs streaming transcription for live audio workflows. Consider Nova-3 for multilingual transcription scenarios or when self-serve terminology customization is important.

Getting Started

Visit the Deepgram website to understand the provider and product context: https://deepgram.com/
Review the Deepgram models and languages documentation for Nova-3: https://developers.deepgram.com/docs/models-languages-overview
Read the official Nova-3 launch page for product background: https://deepgram.com/learn/introducing-nova-3-speech-to-text-api
Test a small batch or streaming transcription workflow before relying on Nova-3 in production.

Key Features

•High-accuracy speech-to-text model
•Supports batch transcription
•Supports streaming transcription
•Multilingual support
•Self-serve terminology customization

Capabilities

•speech-to-text
•batch transcription
•streaming transcription
•multilingual transcription
•terminology customization

Last updated Jun 11, 2026