What is Deepgram?
Deepgram is a voice AI platform offering speech-to-text (STT), text-to-speech (TTS), voice agent and audio intelligence APIs, available in the cloud or self-hosted. Its Flux models are positioned for conversational use: they handle turn-taking, interruptions and conversational context, aimed at live voice agents. Developers can call a single unified Voice Agent API that combines speech recognition, LLM orchestration and speech synthesis, instead of wiring separate components together.
In practice you can transcribe recorded or live audio with Nova-3 and Flux models, including multi-speaker detection and automatic language detection, generate expressive speech with Flux TTS, and analyze conversations for insights. The platform is aimed at contact centers, conversational AI products, speech analytics and medical or podcast transcription, with custom models available for enterprise compliance and niche audio.
Key features
- Real-time speech-to-text with Flux and Nova-3 models
- Text-to-speech with latency as low as 80ms
- Unified Voice Agent API combining STT, LLM orchestration and TTS
- Speaker diarization and automatic language detection
- Audio Intelligence for analysis of conversations
- Cloud or self-hosted deployment
Who is it for
- Developers building voice agents
- Contact centers
- Enterprises with compliance requirements
- Startups and platforms embedding voice features
How to get started with Deepgram
- Create a Deepgram account and claim the 200 USD starting credit.
- Generate an API key in the dashboard and store it securely.
- Send your first REST request or connect via WSS for transcription.
- You are building a voice product or voice agent for developers.
- You need low-latency speech-to-text or text-to-speech.
- You want STT, LLM and TTS combined in one API instead of several components.
- You want a ready-made no-code tool, since the platform has no Polish interface.
- You need a permanent free plan without the starting credit limit.
Pros and cons
- Built for real-time conversation with turn detection and interruption handling
- Single API reduces integration complexity and latency
- Pricing for growth and enterprise plans requires contacting sales
- Free credit on the pay-as-you-go plan is limited to a $200 starting credit
Deepgram pricing
| Plan | Price | Includes |
|---|---|---|
| Pay As You Go | 0.0048 USD / month | Nova-3 Monolingual, streaming, za minutę; bez minimum i bez wygaśnięcia |
| Pay As You Go | 0.0065 USD / month | Flux English, streaming, za minutę |
Prices copied from Deepgram pricing, checked October 12, 2026. Currency as shown by the vendor; taxes may apply.
Is Deepgram free?
There is a free trial. Pay As You Go plan with a $200 free credit, no credit card required, no expiration.
Deepgram alternatives
| Tool | Price from | Free plan | Category |
|---|---|---|---|
 AssemblyAI | not listed | ? | APIs & models |
 Cartesia | 5 USD/mo | yes | APIs & models |
 LocalAI | not listed | yes | APIs & models |
 Lingvanex | not listed | ? | APIs & models |
 Hume AI | not listed | ? | APIs & models |
 Google AI Studio | 0,75 USD | ? | APIs & models |
All alternatives to Deepgram →
FAQ
What is Deepgram used for?
Enterprise voice AI APIs (speech-to-text, text-to-speech, voice agents) for developers and companies building real-time voice applications.
Is Deepgram free?
There is a free trial. Pay As You Go plan with a $200 free credit, no credit card required, no expiration.
How much does Deepgram cost?
The cheapest paid plan is 0.0048 USD/mo (vendor pricing, October 12, 2026).
What are the best Deepgram alternatives?
Closest alternatives: AssemblyAI, Cartesia, LocalAI, Lingvanex, Hume AI.
Data checked October 12, 2026 on the vendor website (deepgram.com).