Deepgram

A speech recognition API built for developers who need speed, accuracy, and scalability. Deepgram's Nova-2 model delivers best-in-class transcription with the lowest word error rates in the industry — handling real-time streaming, phone calls, and multi-speaker audio with ease.

• Nova-2 model — industry-leading accuracy with the lowest word error rates • Real-time streaming — sub-300ms latency for live transcription applications • Speaker diarization — identify and label different speakers in conversations • Language support — 36+ languages with accent and dialect handling • Custom vocabulary — boost accuracy for industry-specific terms and jargon • Audio intelligence — summarization, topic detection, and sentiment analysis

Pay-as-you-go pricing starting at $0.0043/minute. Free tier with $200 credit.