
Deepgram
β Text source-checked Nov 9, 2026Real-time infrastructure for speech-to-text, text-to-speech, and voice agents that listen and respond naturally.
Source-checked against the vendor's site Β· rebuilt from the vendor's own pagesProfile updated Link checked Pricing captured How we rate
In short
Deepgram provides unified APIs for speech-to-text, TTS, and voice agents with flexible deployment options. It offers enterprise-grade security and a freemium model starting with a $200 credit.
Deepgram is a foundational Voice AI platform for developers offering unified APIs for Speech-to-Text, Text-to-Speech, and autonomous Voice Agent orchestration. The platform is available as a managed API with US, EU, and Australia endpoints, in-VPC, or self-hosted.
Speech-to-text, text-to-speech, and voice agent orchestration run on a single enterprise runtime. The Voice Agent API uses a single rate covering speech-to-text, a managed language model, and text-to-speech. Users can bring their own LLM or TTS provider for a lower per-minute rate on the Voice Agent API.
Self-hosted containers run in a user's own VPC or on-premises hardware and require NVIDIA GPUs. Self-hosting is available on the Enterprise plan for teams with data sovereignty or security requirements that rule out cloud processing.
The vendor says that the platform is SOC 2 Type 1 and Type 2 certified with independently audited security controls. The vendor says that Deepgram is HIPAA compliant and signs Business Associate Agreements for Enterprise customers handling electronic Protected Health Information.
According to Deepgram: Deepgram is a foundational Voice AI platform for developers offering unified APIs for Speech-to-Text, Text-to-Speech, and autonomous Voice Agent orchestration. Its pages list this strength: Text-to-speech responses start in as low as 80ms at full 24kHz fidelity. They also state this limitation: Self-hosted deployments require NVIDIA GPUs.
β The AI Tool Bible editorial team
Pros
- β Text-to-speech responses start in as low as 80ms at full 24kHz fidelity.
- β Self-hosted deployment is available in a user's own VPC or on-premises hardware.
- β The Voice Agent API bundles speech-to-text, a managed language model, and text-to-speech into one rate.
Cons
- β οΈ Self-hosted deployments require NVIDIA GPUs.
- β οΈ Concurrency limits apply per project and differ by plan and region.
Pricing (captured )
| Tier | Monthly | Annual | Key features |
|---|---|---|---|
| Pay as you go | β | β |
|
| Growth | β | $4,000+ / year |
|
| Enterprise | β | β |
|
Prices change; we show the grid as captured on Nov 5, 2026. Check current pricing on the vendor's site β
Facts
- Platforms
- api (not source-checked)
- API available
- Yes (not source-checked)
- Free trial
- Yes (not source-checked)
- Company
- Deepgram, Inc. (not source-checked)
- Model used
- Nova, Flux, Speak (proprietary) (not source-checked)
- Hacker News
- 18 stories Β· top 4 points Β· last Oct 29, 2026 Β· checked Nov 10, 2026
- Earliest Wayback capture
- 2016-12-08 (earliest archived capture of the site, not a launch date)
- Headquarters
- United States (Wikidata)
Where this tool sits
Use cases
Frequently asked
- How much does Deepgram cost?
- Deepgram uses a freemium model. You get a free plan with a $200 credit, then pay as you go. A Growth plan with pre-paid credits starts at $4,000+ per year.
- Can I self-host Deepgram?
- Yes, self-hosted containers are available on the Enterprise plan. They run in your VPC or on-premises hardware but require NVIDIA GPUs. This suits teams with strict data sovereignty needs.
- What security certifications does Deepgram have?
- The platform is SOC 2 Type 1 and Type 2 certified. It is also HIPAA compliant and signs Business Associate Agreements for Enterprise customers handling electronic Protected Health Information.
- Can I use my own LLM or TTS provider?
- Yes, users can bring their own LLM or TTS provider. This option allows for a lower per-minute rate on the Voice Agent API compared to using the managed providers.
- What models power Deepgram?
- Deepgram uses proprietary models named Nova, Flux, and Speak. These power its speech-to-text, text-to-speech, and voice agent orchestration capabilities on a single enterprise runtime.