About Deepgram
Deepgram provides APIs for speech recognition, speech synthesis and voice agents. It is built for teams that want to add spoken input and spoken replies to their own products.
What it does
Deepgram lets a program send audio and receive a text transcript, either as a live stream or as a batch job. It also converts text into speech with a set of voices and offers a Voice Agent API. That API handles the back and forth of a phone or chat style conversation, including calls to outside functions such as booking an appointment.
Key features
- Live transcription of streamed audio
- Batch speech-to-text for recorded files
- Text-to-speech with several selectable voices
- Voice agent API with function calling
- Speech output over WebSocket for playback as audio arrives
Who it's for
Developers building voice assistants, automated phone systems, transcription tools or chatbots that need to speak. The site focuses on enterprise voice projects, and smaller teams can also start with the API and an account.
Pricing and good to know
The products are offered as APIs that you call from your own code. Pricing and plan details are on the company's pricing page. Voice work has more moving parts than plain transcription, so test latency and accuracy on your own audio before you commit.
Highlights
- Convert live or recorded audio into text through an API
- Generate spoken audio from text with a range of voices
- Stream speech output over WebSocket into a browser or app
- Build voice agents that call functions during a conversation
- Compare several text-to-speech voices side by side
