Voice Biomarker API Documentation

The Virtuosis voice biomarker API lets healthcare, research and enterprise applications upload speech recordings and retrieve health, wellbeing and communication insights.

Learn more about voice biomarkers and review Virtuosis AI’s clinical validation.

API flow

  1. Create an account for the user.
  2. Upload a Base64-encoded recording, specifying which analysis types to run.
  3. Poll for analysis results until processing completes.

Base URL

https://api.virtuosis.ai/v1.3

What the API covers

  • How to authenticate with the API.
  • How to create accounts, upload recordings, poll analysis results, and inspect usage.
  • Audio requirements for analysis processing.
  • Supported analysis types: wellbeing, Parkinson’s, Alzheimer’s, and communication coach.
  • Shared API error response structure and error types.

Required inputs

InputDescription
API keySent as Authorization: Bearer <API_KEY>.
Audio fileBase64-encoded WAV, MP3, MP4, or OGG recording.
recorded_atISO 8601 timestamp for when the recording began.

Accounts are created under the organisation the API key belongs to and are always assigned the speaker organisation role.

Key facts

  • Public API version: v1.3.
  • API base URL: https://api.virtuosis.ai/v1.3.
  • API authentication: Authorization: Bearer <API_KEY>.
  • Maximum upload size: 50 MB.
  • Recommended analysis polling interval: 15 to 30 seconds.
  • Analysis processing may take up to five minutes.
  • Available analysis types: wellbeing, parkinsons, alzheimers, communication_coach. The analysis array is required and must list at least one type.
  • Usage is metered against credit balances (trial, monthly included, and purchased buckets).

Recording guidance

Recordings should contain at least 30 seconds of free speech. If the recording is not part of a broader conversation, prompt the speaker with a short open-ended question such as:

  • How’s your day going?
  • Can you introduce yourself?

Upload the audio file as a Base64-encoded string in the audio request field. The API does not accept a multipart file upload for this endpoint.

If a clip may contain multiple speakers, set isolate_oldest_speaker to true to diarise the clip and analyse only the speaker estimated to be oldest. The retained speaker’s speech must still meet the 30-second minimum, and recordings longer than 10 minutes are rejected when this option is enabled.

LLM and tool usage

Use the OpenAPI exports when configuring an LLM agent or integration tool:

When implementing from the schema, preserve the three-step flow: create an account, upload a recording, then poll for analysis results. Do not infer additional endpoints, non-production hosts, multipart uploads, or diagnosis workflows.

User-interface disclaimer

Applications that display Virtuosis outputs must include this disclaimer in the user interface:

Not intended to provide a medical diagnosis or to replace clinical judgment. Outputs are for clinical decision support and must be interpreted by a qualified healthcare professional.

Virtuosis provides logo assets in multiple formats and variants, as well as CE marking documentation. See Guidelines and Assets for downloadable assets and implementation requirements.