Deepgram

Overview

Deepgram is a voice AI platform for developers and product teams building speech features and conversational applications. It offers speech-to-text through its Nova models, text-to-speech through Aura, and a Voice Agent API that orchestrates listening, an LLM and speaking for real-time voice bots, all reached through unified APIs. It supports both real-time streaming and batch processing, speech-to-text across 45 or more languages with automatic language detection, and add-ons such as redaction, speaker diarization, keyterm prompting and audio intelligence like summarization and sentiment. It runs in the cloud or self-hosted for in-region and in-house data control, and carries SOC 2 Type 1 and 2, HIPAA, GDPR, CCPA and PCI compliance. Billing is usage-based with no minimums and $200 in free credits to start.

Best for developers building real-time speech-to-text, text-to-speech and voice agents through an API.

Deepgram suits teams building voice into their own product rather than producing audio in a studio: every capability is an endpoint, and the usage-based model rewards steady programmatic volume over occasional creative work. Prototyping costs nothing up front, and self-hosting is what makes it viable where recordings are not allowed to leave the building. Anyone who wants to clone a voice or dub finished video will need a second tool alongside it.

Key Facts

Features

Voice cloning

No, the catalogue is fixed: 40+ prebuilt Aura-2 voices across English, Spanish, Dutch, French, German, Italian and Japanese, with no cloning or custom-voice product on the price list

Real-time use

Yes, real-time streaming speech-to-text and text-to-speech, plus batch processing

Languages

45+ languages for speech-to-text, with automatic language detection

Speech-to-speech and dubbing

No dubbing or translation product. Speech-to-speech runs as a pipeline through the Voice Agent API, which handles listening, thinking and speaking with configurable speech-to-text models, LLM providers and voices, plus function calling mid-conversation

Editing and controls

Voice selection, plus encoding, bit rate, container and sample rate on the output, streamed audio and callbacks. Aura-2 is built to read naturally without SSML markup, and speed and style can be changed mid-conversation without restarting the session

Consent and trust

SOC 2 Type 1 and 2, HIPAA, GDPR, CCPA and PCI compliant; self-hosting for in-house data control

API

Yes, unified APIs for speech-to-text, text-to-speech and voice agents

Integrations

Integrations coming soon.

Deployment and Support

Deployment

• Cloud (SaaS)
• Browser-based
• On-premise

Support

• Community forum

Onboarding

• Documentation and knowledge base

Company size

• Small and mid-sized
• Enterprise

Suitability

The table combines the industries the vendor addresses with the company sizes the tool is aimed at. An empty cell means the combination is not stated, not that it is ruled out.

Pros and Cons

Pricing

• Pay As You Go

  $0.0048

  Per minute of Nova-3 monolingual streaming speech-to-text at the current promotional rate, $0.0077 at list. Nova-3 multilingual $0.0058, Flux English $0.0065. Text-to-speech $0.0150 / 1,000 characters on Aura-1 and $0.030 on Aura-2. Voice Agent API from $0.056 / minute. Add-ons are priced separately: redaction and diarization $0.0020 / minute each, entity detection $0.0017, keyterm prompting $0.0013. No minimums, no expiry, $200 in credits to start without a card.

• Growth

  From $4,000

  Per year in prepaid credits, redeemed against usage at lower rates: Nova-3 monolingual streaming $0.0042 / minute, Aura-1 $0.0135 and Aura-2 $0.027 / 1,000 characters.

• Enterprise

  Not stated

  Custom pricing, volume commitments, self-hosting and dedicated support.

Best For

• Media & Creative
• Technology

Alternatives and Comparisons

What users say

Reviews

Share your experience

Write a Review