Deepgram STT

Deepgram STT delivers ultra-low latency, high-accuracy speech recognition for real-time and pre-recorded audio, powering voice agents and enterprise transcription at scale.

Deepgram STT (Speech-to-Text) is a foundational Voice AI platform, offering APIs for transcription of both live streams and batch files. Its models, like Flux for conversational AI and Nova-3 for high-performance transcription, provide industry-leading accuracy across 36+ languages and dialects. The platform guarantees ultra-low latency (under 300ms) for real-time applications (voice agents, live captions) and includes advanced features: speaker diarization, custom vocabulary, and redaction. Developers can deploy via the cloud API or self-host, ensuring flexibility for various use cases (contact center analytics, media subtitling).

https://deepgram.com/product/speech-to-text-api