Deepgram Speech-to-Text API

What it does

Deepgram's speech-to-text API takes spoken audio, either real-time streams or pre-recorded files, and returns written text. Developers can add transcription, speaker diarization, Smart Formatting, and Keyterm Prompting, with add-on features such as Redaction and Audio Intelligence summaries, topics, sentiment, and intent billed on top.

Teams typically pick a model for the job: Flux STT for real-time voice agents that need turn detection and interruption handling, Nova-3 for general real-time or batch transcription, and industry-specific or custom models for specialized vocabulary. Deployment can run in Deepgram's cloud, through a regional EU endpoint, or self-hosted in your own environment on NVIDIA GPUs.

The API suits developers and engineering teams building voice applications, with enterprise options for HIPAA, SOC 2, GDPR and self-hosting. Access requires creating a Deepgram account and API key; pay-as-you-go includes a $200 credit that does not expire. Pricing varies by model, mode and add-on, and Growth and Enterprise tiers have concurrency and commitment terms to review.

What it can do

• Real-time streaming speech-to-text with ultra-low latency for voice agents
• Pre-recorded audio transcription across 60+ languages
• Language switching and code-switching within utterances
• Custom vocabulary via keyterm prompting and tuned models
• Flexible deployment: Deepgram cloud, regional endpoints, or self-hosted
• Turn detection and natural interruption handling in conversational speech

Example workflows

Real-time voice agent transcription

Illustrative example. Input: live caller audio over a streaming connection. Steps: open a Flux STT stream, receive interim and final transcripts with turn detection, pass final text to an LLM. Output: incremental text transcripts the agent can respond to.

Batch call-recording transcription

Illustrative example. Input: pre-recorded call recording file. Steps: submit the file to Nova-3 with Smart Formatting and speaker diarization, optionally add Redaction. Output: formatted transcript with speaker labels and optional redacted text.

Before you choose

• Speech recognition is metered by audio duration; model choice and optional processing add charges.
• Concurrency limits and prepaid commitments differ by plan.
• Enterprise self-hosting requires suitable infrastructure and a deployment agreement.

Listed under

• Coding
• Call analytics
• Contact centers
• Developers
• Medical transcription
• Speech-to-text
• Transcription

Get started

Sign up, grab an API key, and test models in the Playground.

1. Create a Deepgram account at the sign-up page and generate an API key in the Console.
2. Test models in the Playground before writing integration code.
3. Install an official SDK (Python, JS, Go, .NET, or Java) or call the REST API with your key, selecting Flux STT for real-time or Nova-3 for transcription.

Pay as you go has no minimums and no expiration on the $200 free credit. Growth requires prepaid annual credits starting at $4,000/year. Audio Intelligence and add-ons bill per token or per minute on top of transcription. See the Deepgram pricing page for current rates.

Source: deepgram.com/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Speech-to-text API | Real-time, conversational & accurate | Deepgram ↗deepgram.com
• Deepgram pricing | Scalable speech-to-text, text-to-speech & Voice Agent APIs | Deepgram ↗deepgram.com
• Welcome to Deepgram's docs | Deepgram docs ↗developers.deepgram.com

Company website

deepgram.com: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.