We’ve reinvented Automatic Speech Recognition (ASR) with a complete, deep learning model that allows companies to get faster, more accurate transcription, resulting in more reliable data sets — on-prem or in the cloud. And we do it all with lower hardware and usage costs so we’re a hell of a lot more scalable than big tech players.

From Dark Matter to Deep Learning


Influenced by their background in particle physics, founders Dr. Scott Stephenson and Noah Shutty used AI techniques to find key moments buried within large datasets of recorded audio. After building a speech analysis platform unlike anything available, they set out to build the speech company for the future.


YCombinator invests in Deepgram. Deepgram’s team of PhDs, engineers, and world-class AI scientists releases the world’s first fully deep learning based speech recognition.


Deepgram releases Trained Speech Models to bring companies unparalleled gains in accuracy and speech understanding capabilities. NVIDIA and Slack invest in Deepgram.


Deepgram signs key enterprise customers after expanding offering to include real-time, highly concurrent transcription in the challenging audio environments of phone calls and meetings. Closed A Round.


Deepgram scales engineering, go to market, and operations teams to keep up with skyrocketing demand for its industry leading Enterprise Speech Recognition platform.



Accuracy + Speed = Human Like Conversations

The success of your voicebot is dependent on the accuracy of the transcription to your Conversational AI system. So why accept lower accuracy from a Big Tech Automatic Speech Recognition (ASR) solution. Get 90%+ accuracy and 300-millisecond transcription speed at a fraction of the cost with Deepgram.


Transcribe All Your Calls Without Sacrificing Accuracy, Speed, or Cost

It’s not a pipe dream but a new reality. 90%+ accuracy, 300-millisecond transcription speed, at a fraction of the cost of legacy Automatic Speech Recognition (ASR). Never miss a customer insight, selling opportunity, or coaching moment.


Just OK ASR is Not Good Enough

We get it. Poor transcription accuracy means poor analytics and incomprehensible meaning. You can’t extract insights with 65-80% accuracy. Good news: you don’t have to. Deepgram can achieve 90%+ accuracy and transcribe one hour of audio in one minute at a fraction of the cost of legacy Automatic Speech Recognition (ASR) solutions. We didn’t settle for just ok. You shouldn’t either.

Product Overview

Demo | How Deepgram Achieves Speed & Scalability

Demo | Intro to Batch & Realtime