Speech APIs for transcription, generated voices and live voice agents, billed by usage.
What it does
On this page
Deepgram sells speech infrastructure: transcription, generated speech and a voice agent API, each reachable over an HTTP request or a streaming socket.
Who it is for
Developers and product teams wiring voice into agents, contact centres, meeting tools and assistants, where latency and cost per minute are the real constraints.
How it works
Audio streams into the transcription endpoint and text comes back while the speaker is still talking, which is what makes a spoken reply feel immediate. Its speech generation model, Flux, is aimed at conversation rather than narration; the site puts the catalogue at 45-plus languages and more than 100 voice profiles. A voice agent endpoint adds turn-taking, interruption handling and function calling, so a caller can complete a booking mid-call. A speech analytics product reads the same audio for patterns.
Pricing
Usage is metered. A new account starts with a $200 credit and no card, then pays as it goes. Growth sells prepaid credits for a year and advertises savings up to 20%. Enterprise is quoted for high volume, deployment or support needs.
Notable limits
Everything is priced by consumption, so cost tracks minutes and characters rather than seats, and a long-running agent needs its own budget watch.
Source: https://deepgram.com/