Deepgram

Founded by Scott Stephenson in 2015

Convert speech to text and generate human-like spoken voice with real-time APIs

Pricing model

Free

Rating

★ People love it

Time to value

Quick Setup (< 1 hour)

You can use Deepgram to convert spoken audio into text and turn written text into lifelike speech through fast APIs. The system transcribes live streams or recorded audio files with high accuracy, handling accents, background noise, and specialized terminology. You can also build conversational voice agents that understand user interruptions and respond in real time with minimal delay. Deepgram connects with language models to drive automated customer support calls, generate meeting transcripts, extract key conversational insights, and run voice applications either in the cloud or on self-hosted infrastructure.

What Deepgram does

Stream live audio for instant speech-to-text transcriptionConvert written text into natural spoken audioConnect speech input and voice output directly to language modelsDetect conversation turns and handle speaker interruptionsTranscribe pre-recorded audio files in batch modeTrigger external API calls during active voice conversationsDeploy speech models directly onto self-hosted hardwareReal-time streaming speech-to-text transcription with low latencyExpressive text-to-speech voice generation operating under 100msNative turn-taking detection and interruption handling for voice agentsSingle unified API combining speech recognition, text generation, and speech synthesisDeployment options across managed cloud, private VPC, and on-premises environmentsAudio analysis tools for conversational topic extraction and summarizationBuilt-in function calling and context injection during active calls