From $16 per month🇺🇸
Video and audio editor where you cut through the transcript, with the Underlord AI co-editor, Studio Sound and translation.
Deepgram AI features, pricing and GDPR (October 2026)
Voice API for transcription, text to speech and voice agents, billed per minute, in the cloud or self-hosted.
Usage-basedFree trial
from $0.00 usage-based
Pricing details
Pay As You Go with no minimum and $200 starting credit. Nova-3 monolingual $0.0043 per minute for pre-recorded audio, streaming currently $0.0048 as a limited-time promotional rate, Aura-1 $0.015 per 1,000 characters, Voice Agent API from $0.075 per minute. Growth with prepaid credits from $4,000 per year, as of September 2026.
What is Deepgram?
Deepgram is a developer platform for speech to text, text to speech and voice agents. Product and engineering teams use it to add transcription or spoken conversations to apps, contact centers and meeting tools, paying per minute or per character.
The San Francisco company offers the Nova and Flux speech recognition models, Aura and Flux voices, a Voice Agent API that orchestrates recognition, a language model and speech output, and Audio Intelligence for conversation analytics. According to the documentation, Nova-3 covers more than 50 languages and has a multilingual mode for ten languages, including English, German, Spanish and Hindi.
Deployment options include Deepgram's cloud and self-hosted installations. The vendor names Twilio, Cloudflare and IBM among its customers and targets contact centers, conversational AI, podcasts and medical transcription.
Deepgram AI features and use cases
- Transcribes calls in real timeNova-3 converts streamed audio into text while a call or meeting is running and can also process stored files. Product and support teams feed the text into search, quality checks or live assistance without waiting for manual notes.
- Runs spoken customer conversationsThe Voice Agent API chains speech recognition, a language model and speech output in one call and handles turn-taking and interruptions. Product teams can build phone or app assistants without stitching three vendors together.
- Generates low-latency speechAura and Flux voices turn text into speech quickly enough for live dialogue, with latency from 80 milliseconds according to the vendor. This keeps voice interfaces responsive for callers and app users.
- Summarizes conversationsAudio Intelligence produces summaries of transcribed conversations and bills them by input and output tokens. Customer experience and product teams get a short account of each call and can spot recurring issues without reading full transcripts.
What sets Deepgram apart
- One API covers recognition, speech output and agent orchestration, each priced per minute or per character.
- Self-hosted deployment is available in addition to the cloud API.
- Pay As You Go needs no credit card and starts with $200 in credit.
Where Deepgram is strongest
United States Available internationally
Deepgram, Inc. is headquartered in San Francisco. Processing runs in the US by default, and a dedicated EU endpoint keeps processing within the EU.
Questions to ask before you buy Deepgram
Deepgram processes data in the US by default. A dedicated EU endpoint keeps processing within the EU, and self-hosting is a further option, so review the DPA and transfer terms for the setup you choose. Recording and transcribing calls requires notice or consent under state wiretap laws such as California's two-party consent rule. Costs scale with minutes and characters; model concurrency and volume before committing to Growth.
Using Deepgram in the EU: GDPR and AI Act
- Data locationMultiple regions, EU region available (according to the vendor). Hosting regions: United States, EU, Australia, India.
- Data processing agreementNot published by the vendor: ask for one (Art. 28 GDPR).
- EU AI ActCheck whether the AI prepares decisions about people. If not, the high-risk rules usually do not apply. The high-risk obligations apply from December 2, 2027 (Digital Omnibus).
General information with sources, not legal advice.
Deepgram reviews
No reviews yet. Be the first to share how your team uses it.
Every review shows its verification level. We read each one before it goes live. How we check reviews
Questions about Deepgram
How much does Deepgram transcription cost?
On Pay As You Go, Nova-3 monolingual costs $0.0043 per minute for pre-recorded audio. Streaming currently costs $0.0048 per minute as a limited-time promotional rate instead of the regular $0.0077. Growth customers prepay at least $4,000 per year and save up to 20 percent, as of September 2026.
Which languages does Nova-3 support?
The documentation lists multilingual support for English, Spanish, French, German, Hindi, Russian, Portuguese, Japanese, Italian and Dutch, with more than 70 language codes including regional variants.
Can Deepgram be self-hosted?
Yes. The vendor offers self-hosted deployment alongside its cloud API, which also changes where audio data is processed.
Deepgram update history1 update
- Deepgram offers regional endpoints in Australia and India in addition to the US and EU. Source (opens in a new tab)
Changes the vendor made and confirmed user reports, each checked against the vendor's own pages. We review every tool monthly.
Alternatives to Deepgram
Tools in the same category (Transcription) with the most overlapping markets.
From $6 per month🇺🇸
Voice AI platform for text to speech, voice cloning, transcription, dubbing and music, as a web app and via API.
Free🇺🇸
Mozilla.ai project that packs a language model and its runtime into a single executable file with no installation.
Research paper🇺🇸
Microsoft report on VibeVoice, a model for long-form speech synthesis with up to four speakers, since extended to speech recognition.