Nova-3, Explained Simply
What Nova-3 is, what it does well, and what it costs — released 2025-02-12 by Deepgram.
- Deepgram just released Nova-3, their smartest AI yet for turning spoken audio into written text.
- It is designed to be highly accurate, even when people talk fast, have accents, or use tricky vocabulary.
- It is built mainly for developers who want to add voice-to-text features to their own apps.
Want a heads-up when Deepgram updates?
We email you the moment a new Deepgram version drops — plus plain-English release notes like these. No spam.
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Nova-3 is a speech-to-text AI model made by a company called Deepgram. "Speech-to-text" simply means it listens to audio recordings and types out exactly what is being said. This is the kind of technology that powers live captions on videos or voice assistants on your phone. The people who will want this the most are app developers and software engineers who need a reliable way to convert human speech into written words for their projects.
What it does well
- It is highly accurate at understanding spoken words and turning them into text.
- It handles "hard" audio well, meaning it can figure out what people are saying even if they mumble, talk over each other, or have background noise.
- It is good at understanding different accents and tricky vocabulary, so it does not get easily confused by unusual words or names.
This is our plain-language summary. Read the complete, official notes on the Deepgram official changelog ↗.
Google Calendar