What is Deepgram?
Deepgram is a developer voice AI platform offering APIs for real-time and batch speech-to-text (Flux and Nova-3 models), text-to-speech (Flux TTS and Aura-2), audio intelligence, and a Voice Agent API. It can run in Deepgram's cloud or self-hosted, and new accounts get $200 in free credit.
Building a voice app used to mean stitching together a transcription service, a language model and a speech engine, then fighting latency and interruptions. Deepgram offers each layer as an API and also a single Voice Agent API that handles turn-taking, interruptions and context in one call. Pricing is per minute or per character, and teams with strict data rules can self-host instead of sending audio to a shared cloud.
Main Features
- Transcribes streaming or prerecorded audio with the Flux and Nova-3 speech-to-text models.
- Generates speech with Flux TTS and Aura-2 voices, with latency as low as 80 ms.
- Runs full voice agents that handle turn-taking and interruptions through one Voice Agent API.
- Adds speaker diarization, smart formatting and entity detection to transcripts.
- Provides an Audio Intelligence API for insights from conversations.
- Supports 45+ languages for transcription.
- Deploys in Deepgram's cloud or self-hosted.
- Holds SOC 2 Type 1 and Type 2 certification with HIPAA, GDPR and CCPA options.
How to Get Started with Deepgram
- Create a free account on deepgram.com to receive $200 in credit, no card required.
- Generate an API key in the console.
- Send an audio file or stream to the speech-to-text API, or text to the TTS API, using an SDK or REST.
- Scale on pay as you go or move to a Growth or Enterprise plan.
Deepgram Pricing
New Pay As You Go accounts get $200 in free credit with no minimums. After that, Flux English streaming is about $0.0065 per minute, Aura-2 TTS $0.030 per 1,000 characters, and the Voice Agent API starts at $0.075 per minute.
Who Should Use It?
- Developers adding transcription or voice to apps
- Contact centers analyzing and automating calls
- Companies building real-time voice agents
- Media teams generating captions and transcripts at scale
- Regulated industries that need self-hosted speech AI
Deepgram Alternatives
- Resemble AI: focused on voice cloning and deepfake detection
- Murf AI: a no-code voiceover studio instead of developer APIs
- CAMB AI: built for dubbing and translating media
Frequently Asked Questions
Is Deepgram free?
New accounts get $200 in free credit on Pay As You Go, with no card and no expiration. After the credit is used, you pay per minute of audio or per 1,000 characters of speech, with discounts on the Growth plan.
What models does Deepgram offer?
Its pricing page lists Flux and Nova-3 for speech-to-text, and Flux TTS and Aura-2 for text-to-speech, plus a Voice Agent API that combines listening, reasoning and speaking in one call.
Can Deepgram be self-hosted?
Yes. Deepgram can run in its own cloud or in your environment, and it lists SOC 2 certification plus HIPAA, GDPR and CCPA compliance options for sensitive audio.