Deepgram
used in
0
stacks
Voice AI platform providing speech-to-text, text-to-speech, and a unified voice agent API.
Offers the Nova family of speech-to-text models, the Flux model for conversational turn-taking, and Aura text-to-speech, available via realtime streaming and batch APIs. A Voice Agent API combines STT, LLM orchestration, and TTS behind a single interface. Supports cloud and self-hosted deployment, with usage-based pricing and free starter credits. Commonly used as the transcription layer inside voice agent frameworks.
Strengths
- Low-latency streaming STT widely used in realtime voice agent stacks
- Unified Voice Agent API combining STT, LLM orchestration, and TTS
- Cloud and self-hosted deployment options
- Usage-based pricing with free credits to start
Tradeoffs
- Proprietary models create vendor lock-in compared with open-weight options
- Accuracy on niche accents, domains, and lower-resource languages can require model tuning or keyword boosting