Deepgram is a voice AI company that sells speech-to-text, text-to-speech and voice agent APIs; an API is an interface that another program calls over the internet. Its Nova speech-to-text models support more than 45 languages, and its Flux models are built for live conversations with voice agents.
A developer sends a pre-recorded audio file or a live audio stream to Deepgram and receives the transcript. Options include speaker diarization (labels that show which speaker said each part), smart formatting for readability and automatic language detection. The text-to-speech models do the reverse and turn text into spoken audio. The Voice Agent API combines speech-to-text, a language model and text-to-speech in one API for real-time voice conversations. SDKs exist for Python, JavaScript, Go, .NET and Java.
Deepgram runs in its own cloud or self-hosted on the customer's servers. The cloud API is billed by usage, per minute of audio, with free credit at sign-up and prepaid yearly plans for larger use.