Deepgram Nova
NewDeepgram's most accurate and fastest speech-to-text model for production applications.
About Deepgram Nova
Deepgram Nova-2 is Deepgram's flagship speech recognition model, delivering best-in-class accuracy at speeds 30x faster than real-time with the lowest latency in the industry—making it ideal for real-time voice AI agents, live captioning, and call center analytics. The Nova model family supports 35+ languages, speaker diarization, smart formatting, and custom vocabulary, and can process audio through both batch API and streaming WebSocket connections. Nova-2 is used by companies like NASA, Spotify, and Twilio to power voice interfaces where speed and accuracy are both critical.
Pros
- 30x faster than real-time with industry-leading low latency
- Streaming WebSocket API ideal for real-time voice applications
- Best-in-class accuracy with Nova-2 architecture
Cons
- deepgram-ai already exists; this covers Nova specifically
- Pricing can grow quickly for high-volume telephony applications
Related Tools
Ultra-realistic AI voice generation and cloning API
Industry-leading AI voice cloning that replicates any voice with exceptional naturalness and emotional range.
OpenAI's state-of-the-art speech recognition API for transcription and translation.