API directory · Speech-to-text API (medical model line)
Deepgram API
Fully public speech-to-text API with a medical-tuned Nova-3 model selected by parameter. Endpoints, concurrency limits and per-minute prices are all published.
Short answer
What does the Deepgram API give a developer?
Fully public speech-to-text API with a medical-tuned Nova-3 model selected by parameter. Endpoints, concurrency limits and per-minute prices are all published.
Base URL: https://api.deepgram.com (EU: api.eu.deepgram.com · AU: api.au.deepgram.com)
Auth: Authorization: Token <API_KEY>
Cite as: Deepgram API listing, Compare Healthcare API, last reviewed 2026-09-01.
Endpoints
Copied from the vendor's own reference. Where a path is described rather than published, the description is what the vendor states.
| Method | Endpoint | Mode | Purpose |
|---|---|---|---|
| POST | /v1/listen?model=nova-3-medical | Batch | Transcribe pre-recorded audio with the medical model; diarisation and keyterm prompting are query parameters. |
| WS | wss://api.deepgram.com/v1/listen | Streaming | Real-time transcription over WebSocket with interim results. |
Published rate limits
Numbers the vendor states publicly. Anything unpublished should be made contractual before you plan capacity.
| Pay-as-you-go, pre-recorded | Up to 50 concurrent requests per project (Nova-3) |
|---|---|
| Pay-as-you-go, streaming | Up to 150 concurrent streams per project (Nova-3) |
| Secondary self-serve projects | 1 concurrent stream |
| Per-minute request caps | None documented — limits are concurrency-based; expect HTTP 429 on breach |
Published pricing
List prices as published on the vendor's own pricing page at the verification date. Follow the source link before you quote these to anyone.
| Nova-3 monolingual, streaming | $0.0077/min list (promotional $0.0048/min at time of verification) |
|---|---|
| Nova-3 monolingual, batch | $0.0043/min pay-as-you-go |
| Nova-3 Medical | Not itemised separately on the pricing page — confirm the medical model rate with the vendor |
| Speaker diarisation add-on | $0.0020/min |
| Keyterm prompting add-on | $0.0013/min |
| Free credit | $200, no credit card required |
Data handling and compliance
The three answers a health-system security review will ask you for, as published.
| BAA | Available; compliance documentation is requested through an account executive |
|---|---|
| Retention | Configurable; documented in the trust and security pages |
| Training on PHI | Enterprise data-handling terms — get the model-improvement position in writing |
How to integrate the Deepgram API
An ordered path based on the public documentation, written for an engineering team scoping the work.
- 01
Get a key
Sign up, create a project key, and keep keys server-side; a browser-exposed key is a PHI incident waiting to happen.
- 02
Transcribe one file
POST a recorded encounter to /v1/listen with model=nova-3-medical and diarize=true, then read the word-level output.
- 03
Add clinical vocabulary
Use keyterm prompting for drug names, local abbreviations and specialty jargon rather than accepting baseline output.
- 04
Move to streaming
Open the WebSocket for live encounters and handle interim versus final results explicitly in your UI.
- 05
Build the note layer
This is where the real project starts: prompting, template adherence, hallucination controls and evaluation are all yours.
- 06
Plan concurrency
Model peak concurrent encounters against the documented 50/150 limits and request an increase before launch, not during it.
What the documentation does not warn you about
The judgement an endpoint list cannot give you.
Excellent transcription infrastructure and the wrong purchase if you expected a note. Everything between transcript and chart-ready documentation becomes your permanent engineering scope — including clinical safety.
Primary sources for this listing
Every fact above traces to one of these pages. If one has changed, tell us and the listing is re-verified.
- API rate limits https://developers.deepgram.com/reference/api-rate-limits
- Authentication https://developers.deepgram.com/reference/authentication
- Nova-3 Medical announcement https://deepgram.com/learn/introducing-nova-3-medical-speech-to-text-api
- Pricing https://deepgram.com/pricing
- Data privacy & compliance https://developers.deepgram.com/trust-security/data-privacy-compliance
Evidence & sources
Standards the integration steps above assume. Vendor-specific evidence is the primary source list on this page.
U.S. Department of Health & Human Services · Regulation · Tier A — primary documentation
Supports: The contractual clauses a documentation vendor's BAA must contain.
U.S. Department of Health & Human Services · Regulation · Tier A — primary documentation
Supports: Administrative, physical and technical safeguards a vendor handling recorded encounter audio must implement.
HL7 International · Standard · Tier A — primary documentation
Supports: Resource definitions (DocumentReference, Composition, Encounter, Condition, MedicationRequest) that clinical documentation output must map onto.
Source tiers are defined on the methodology page. Outbound links are unaffiliated and carry no commercial relationship.
Frequently asked questions
- Does Deepgram have a public API?
- Fully public speech-to-text API with a medical-tuned Nova-3 model selected by parameter. Endpoints, concurrency limits and per-minute prices are all published.
- How do you authenticate with the Deepgram API?
- Authorization: Token <API_KEY>
- What are the Deepgram API rate limits?
- Pay-as-you-go, pre-recorded: Up to 50 concurrent requests per project (Nova-3). Pay-as-you-go, streaming: Up to 150 concurrent streams per project (Nova-3). Secondary self-serve projects: 1 concurrent stream. Per-minute request caps: None documented — limits are concurrency-based; expect HTTP 429 on breach
- How much does the Deepgram API cost?
- Nova-3 monolingual, streaming: $0.0077/min list (promotional $0.0048/min at time of verification). Nova-3 monolingual, batch: $0.0043/min pay-as-you-go. Nova-3 Medical: Not itemised separately on the pricing page — confirm the medical model rate with the vendor. Speaker diarisation add-on: $0.0020/min. Keyterm prompting add-on: $0.0013/min. Free credit: $200, no credit card required
- Can you evaluate Deepgram without a sales conversation?
- Yes — self-serve key, $200 credit, published prices; higher concurrency requires sales