Deepgram Nova-3 Transcription and Intent
Deepgram Nova-3 transcribes recorded audio with timestamps and diarization, and its Pro tier adds summarization, topic, sentiment and intent analysis.
- Deepgram
- 2 variants
- Price
- See price
- Speed
- See speed
- Best for
- Voice to Text
- Price
- See price
- Speed
- ~10s
- Best for
- Voice to Text
Working with Deepgram Nova-3.
Nova-3 is a speech-to-text family with an analytics tier. The base model transcribes pre-recorded audio with word-level timestamps, speaker diarization and automatic language detection; Pro adds summarization, topic and entity detection, sentiment and intent analysis and smart formatting on top of the transcript.
When to choose Deepgram Nova-3
Choose Nova-3 Pro when you want understanding as well as text — call analytics, support triage, meeting summaries. Whisper is the pick for raw multilingual transcription at speed.
Choosing between the variants
The base Speech to Text variant is the transcription workhorse with timestamps and diarization. Pro is the one that also returns summaries, topics, entities, sentiment and intent, so a pipeline needs fewer follow-up calls.
Questions about Deepgram Nova-3
The short answers. For anything specific to your setup, talk to our team.
Talk to an engineerPlan for your expected volume.Summarization, topic and entity detection, sentiment and intent analysis, and smart formatting — analysis on top of the transcript.
Yes, automatic language detection is built into the base model along with diarization and word-level timestamps.
The variants here target pre-recorded audio. For long archives at speed, Incredibly Fast Whisper is worth comparing.