vocametrix-python

Official Python SDK for the Vocametrix API — voice analysis, speech therapy, and acoustic measurement for speech-language pathologists, voice researchers, and developers.

Get an API key

Vocametrix is a commercial API. You need an account to use this SDK.

• Sign up — create an account and get your API key. New accounts include a free trial (5 minutes of analysis or 5 API credits, whichever comes first) so you can run the quickstart below without entering payment details.
• Pricing — see plans and rates once your trial is used up.

Once you have a key, every call is one VOCAMETRIX_API_KEY env var away.

Install

pip install vocametrix

Requires Python ≥ 3.9.

30-second quickstart

export VOCAMETRIX_API_KEY="your-api-key-here"

from vocametrix import VocametrixClient

client = VocametrixClient()  # reads VOCAMETRIX_API_KEY from environment

# AVQI — clinically validated dysphonia score (Maryn & Weenink 2015)
result = client.avqi.calculate(sustained_vowel="vowel.wav")
print(result["AVQI"])    # e.g. 1.8 → normal (threshold: 2.97)
print(result["CPP"], result["HNR25"])

Authentication

Set the VOCAMETRIX_API_KEY environment variable, or pass it explicitly:

from vocametrix import VocametrixClient

# From env var (recommended — never hardcode keys)
client = VocametrixClient()

# Explicit key (e.g. loaded from a secrets manager)
client = VocametrixClient(api_key="va_...")

Never hardcode API keys in source code. Use environment variables or a secrets manager.

What's included

Workflows

Pronunciation assessment

result = client.pronunciation.assess(
    audio="recording.wav",
    reference_text="Hello, my name is Alex.",
    locale="en-US",
)
print(result["accuracyScore"], result["fluencyScore"])
for word in result.get("words", []):
    print(word["word"], word["accuracyScore"])

Batch AVQI over a folder

import pathlib

for wav in pathlib.Path("./recordings").glob("*.wav"):
    result = client.avqi.calculate(sustained_vowel=str(wav))
    avqi = result["AVQI"]
    label = "Normal" if avqi < 2.97 else "Dysphonic"
    print(f"{wav.name}: AVQI={avqi:.2f}  ({label})")

Async transcription with SSE

for event in client.transcription.stream("recording.wav", locale="en-US"):
    print(event.status, event.progress)
    if event.is_terminal_success:
        print("Transcript:", event.display_text)

Async client

import asyncio
from vocametrix import AsyncVocametrixClient

async def main():
    async with AsyncVocametrixClient() as client:
        result = await client._aget(
            "https://platform.vocametrix.com/api/calculate-avqi",
            params={"svFileId": "..."},
        )
        print(result)

asyncio.run(main())

Error handling

from vocametrix.exceptions import (
    VocametrixAuthError,       # 401 — bad or missing API key
    VocametrixRateLimitError,  # 429 — SDK retries automatically (up to 3×)
    VocametrixValidationError, # 422 — invalid parameter values
    VocametrixServerError,     # 5xx — SDK retries automatically
)

try:
    result = client.avqi.calculate(sustained_vowel="vowel.wav")
except VocametrixRateLimitError as e:
    print(f"Rate limited. Retry after {e.retry_after}s")
except VocametrixAuthError:
    print("Check your API key at https://www.vocametrix.com/registration")

The SDK retries 429, 500, 502, 503, 504 with exponential backoff (up to 3 retries). Non-retriable errors (4xx except 429) raise immediately.

Longer examples

See vocametrix-examples for:

• End-to-end workflow scripts (batch pronunciation, full voice assessment, prosody loops)
• Jupyter notebooks with cohort visualizations

API reference

• Interactive docs
• OpenAPI 3.1 spec — for typed client generation or Swagger UI

Related projects

The Vocametrix ecosystem:

• 📘 Vocametrix API documentation — full reference for all 48 endpoints (clinical voice quality, pronunciation assessment, speech-to-text, phoneme detection, prosody similarity, eGeMAPS feature extraction, and more).
• 📐 OpenAPI 3.1 specification — machine-readable API schema. Generate clients in any language.
• 🟦 vocametrix-js — official TypeScript / JavaScript SDK (npm install vocametrix).
• 🤖 vocametrix-mcp — official MCP server for Claude, Cursor, Zed, Windsurf, and any MCP-compatible agent.

Contributing

git clone https://github.com/pmarmaroli/vocametrix-python
cd vocametrix-python
pip install -e ".[dev]"

# Regenerate the low-level client from the live OpenAPI spec
python tasks.py regenerate

# Run unit tests (no API key needed — all mocked)
pytest tests/unit/

# Run integration tests (real API calls — requires VOCAMETRIX_TEST_API_KEY)
pytest tests/integration/

The src/vocametrix/_generated/ directory is auto-generated. Do not edit it manually. All hand-written logic lives in src/vocametrix/client.py, _namespaces.py, _http.py, and exceptions.py.

License

The vocametrix SDK is released under the MIT License. You're free to use, modify, and redistribute the SDK source code.

Calling the Vocametrix API with this SDK requires an API key, which is issued with a paid Vocametrix account. The SDK code is free; the service is not. See pricing.