Backed By Deepgram
Fast and accurate speech recognition and voice agent.
A production-grade voice agent API enabling real‑time and batch transcription, speaker diarization, emotion analysis, text‑to‑speech, and conversational voice agent workflows. Widely used in call centers, contact centers, and automated voice bots.
Live call transcription
Emotion and sentiment analysis
Voice agent for IVR automation
Contact center conversation intelligence
$200 in free credits (approx. 40 hours of voice agent usage)
Includes STT, TTS, and LLM orchestration (full Deepgram stack)
Built-in rate reductions for bringing your own LLM or TTS models
Unified Real-Time API (STT + LLM + TTS orchestration)
Bring Your Own Model (BYOM) support for LLM and TTS
Real-time, low-latency performance
Multilingual support (100+ languages supported by Deepgram's models like Nova-3 Multilingual)
Speaker Diarization
Real user experiences from across different platforms
Shortcut is built entirely on Deepgram's Voice Agent API and speech models! Deepgram's APIs were a crucial part of what makes Shortcut work so seamlessly. The accuracy and speed of their models have been game-changing for Shortcut.
Sharon Yeh (used to build Shortcut)
12mo ago
re low latency and high reliability (99.9%+ uptime).
egration complexity and cost/hour.
our flat rate may be cost-prohibitive compared to pay-per-use alternatives.
age support not covered by Deepgram's current models.
Developer simplicity (unified API, no black box limitations) combined with enterprise control, flexibility (BYOM/BYO TTS/LLM), high accuracy, and cost-effectiveness at massive scale.
Higher cost for smaller teams/projects compared to some competitors Occasional TTS quirks reported by some users Variable accuracy reported by some users (though overall accuracy is highly praised)