What is Deepgram?
Deepgram is an enterprise voice AI platform that provides speech-to-text, text-to-speech, and voice agent APIs for developers, platforms, and enterprises building real-time conversational applications. Its differentiator is a unified Voice Agent API that combines speech recognition, speech synthesis, and LLM orchestration in a single API, built on conversation-aware models designed to handle turn-taking, interruptions, and conversational context.
Key Features
- Flux STT (Speech-to-Text): Real-time transcription with production accuracy, built to detect end-of-turn and interruptions and stream transcripts in real time. It is available in real time or batch, in the cloud or self-hosted.
- Flux TTS (Text-to-Speech): Expressive voice synthesis with latency as low as 80ms, designed for live conversations. Deepgram describes Flux TTS as the first voice model that holds tone, expression, and continuity across an entire conversation.
- Unified Voice Agent API: Combines speech-to-text, text-to-speech, and LLM orchestration into a single API call instead of stitching together separate components, reducing complexity, latency, and cost.
- LLM Orchestration: Coordinates context, memory, and AI reasoning, including conversation history and system settings, system and prompt updates, response injections, and function calling to execute actions or fetch data.
- Audio Intelligence: Extracts insight from every conversation.
- Transport Layer Support: A partner-provided transport layer bridges streamed audio into Deepgram's APIs and plays back TTS audio to end users, including telephony partners via PSTN/SIP.
- Business Logic and External Systems Integration: Custom application logic executes domain-specific actions, integrates external APIs, or adds validation, while external databases, APIs, and back-office systems feed results back into the conversation context for grounding.
- Deployment Flexibility: Available in real time or batch, in the cloud or self-hosted, with custom models available for enterprises with unique workflows and compliance needs.
Use Cases
- Voice agent developers: Teams building conversational voice agents can use the unified Voice Agent API to handle speech recognition, LLM reasoning, and speech synthesis in one pipeline rather than integrating separate vendors.
- Platforms and partners embedding voice AI: Platforms can integrate enterprise-grade voice AI through Deepgram's partner program, including telephony transport via PSTN/SIP.
- Real-time and batch transcription workflows: Product teams can run real-time transcription for live audio or batch transcription for recorded audio, in the cloud or self-hosted.
- Conversational TTS applications: Developers building live voice experiences can use Flux TTS to maintain tone, expression, and continuity across a conversation.
- Enterprises with compliance requirements: Organizations with unique workflows and compliance needs can pursue custom models and self-hosted deployment through Deepgram's sales process.
Who It Is For
Deepgram targets developers and product teams building with APIs, platforms and partners embedding voice AI, and enterprises with custom workflow or compliance requirements. It solves the problem of assembling a voice agent from separate speech-to-text, text-to-speech, and LLM components by unifying them into a single API, while offering cloud or self-hosted deployment for teams with infrastructure or compliance constraints.
Pricing
The source does not publish a pricing model or rate card. Deepgram offers a free trial ("Try for free") and a promotional offer of $1 back on every $1 spent on Flux TTS, up to $500; Flux TTS is also free to build with through September 12th, with up to 45 concurrent streaming connections in the US and 5 in the EU and AU. Enterprises and custom-model buyers are directed to talk to sales.
Frequently Asked Questions
What is Deepgram used for?
Deepgram is used to build voice AI applications: real-time and batch speech-to-text, expressive text-to-speech, and full voice agents that combine STT, TTS, and LLM orchestration through a single API. It also provides audio intelligence to extract insight from conversations.
Is Deepgram free?
Deepgram offers a free trial, and Flux TTS can be built with free through September 12th with up to 45 concurrent streaming connections in the US and 5 in the EU and AU. A promotional offer also returns $1 in credit for every $1 spent on Flux TTS, up to $500.
What makes Deepgram's Voice Agent API different?
Deepgram's Voice Agent API unifies speech-to-text, text-to-speech, and LLM orchestration into a single API instead of requiring separate components to be stitched together. According to Deepgram, this reduces complexity, latency, and cost.
Can Deepgram be self-hosted?
Yes. Deepgram states its voice AI is available in real time or batch, in the cloud or self-hosted. Enterprises with unique workflows and compliance needs can also discuss custom models with Deepgram's sales team.
Who should use Deepgram?
Deepgram is aimed at developers and product teams building voice features via APIs, platforms and partners embedding voice AI (including telephony via PSTN/SIP), and enterprises needing custom models or compliance-oriented deployment. Teams that want one vendor for STT, TTS, and agent orchestration are the clearest fit.
