Deepgram
4.5
Deepgram provides speech-to-text, text-to-speech, voice agent and audio intelligence APIs. Pre-recorded Nova-3 monolingual is $0.0043 a minute and multilingual $0.0052; streaming runs $0.0077 to $0.0092. New accounts get a $200 credit that does not expire, with no card required. It publishes SOC 2, HIPAA, GDPR and PCI compliance and offers self-hosted deployment on enterprise.
- Commercial use OK
- Free tier
- Enterprise contract
- SOC 2
- Pricing
- 従量課金
- Free tier
- Free tier
- Commercial use
- Permitted
- Japanese UI
- Not confirmed
- Enterprise contract
- Available
- Compliance
- SOC 2
What it is good for
- High-volume transcription where per-minute cost dominates
- Real-time streaming transcription for live applications
- Regulated environments needing self-hosted deployment
- Building voice agents on a single vendor's stack
Strengths and trade-offs
Strengths
- Among the lowest published per-minute rates for pre-recorded audio
- $200 non-expiring credit is a generous evaluation allowance
- Self-hosted option for data that cannot leave your environment
- Broad compliance coverage published
Trade-offs
- No ongoing free tier after the initial credit
- Japanese accuracy should be verified on your own audio
- Growth tier requires a $4,000 annual prepayment
Editors' review
Price per minute is the whole decision at volume
At a million minutes a year, the gap between $0.0043 and $0.0077 a minute is thousands of dollars. For low volume the difference is irrelevant and convenience wins.
The $200 credit with no expiry is enough to run a genuine accuracy comparison on your own Japanese audio, which is the only test that matters.
Verdict
The cost leader for bulk pre-recorded transcription. Test Japanese accuracy with the free credit.
Compiled from official documentation and public sources, with our own editorial judgement. We do not publish figures we cannot source.