Soniox Python SDK

Official Python SDK for Soniox Speech-to-Text API. Built with httpx for both synchronous and asynchronous support.

Features

• 🎯 Complete API Coverage: Full support for Soniox REST API
• ⚡ Async & Sync: Full support for both synchronous and asynchronous operations
• 🔒 Type Safe: Built with Pydantic v2 for robust type checking and validation
• 📝 Comprehensive Logging: Built-in logging with the soniox logger
• 🌍 60+ Languages: Transcribe speech in multiple languages with language hints
• 🎭 Speaker Diarization: Identify different speakers in audio
• 🔍 Language Identification: Automatic language detection
• 📊 Word-Level Timestamps: Get precise timing for each word
• 🎯 Context Support: Improve accuracy with domain-specific context

Installation

pip install soniox

Quick Start

Authentication

Set your API key as an environment variable:

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

Or pass it directly when initializing the client:

from soniox import SonioxClient

client = SonioxClient(api_key="your-api-key-here")

Basic Usage

Transcribe an Audio File

Synchronous:

import time
from soniox import SonioxClient

client = SonioxClient()

# Submit transcription job
job = client.transcribe_file("path/to/audio.wav")
print(f"Job ID: {job.id}")
print(f"Status: {job.status}")

# Poll for completion
while True:
    job = client.get_transcription_job(job.id)
    if job.status == "completed":
        break
    time.sleep(1)

# Get the transcript
result = client.get_transcription_result(job.id)
print(f"Transcript: {result.text}")
print(f"Tokens: {len(result.tokens)}")

Asynchronous:

import asyncio
from soniox import SonioxClient

async def transcribe():
    client = SonioxClient()
    
    # Submit transcription job
    job = await client.transcribe_file_async("path/to/audio.wav")
    print(f"Job ID: {job.id}")
    
    # Poll for completion
    while True:
        job = await client.get_transcription_job_async(job.id)
        if job.status == "completed":
            break
        await asyncio.sleep(1)
    
    # Get the transcript
    result = await client.get_transcription_result_async(job.id)
    print(f"Transcript: {result.text}")

asyncio.run(transcribe())

Transcribe with Custom Configuration

You can pass configuration options either as a TranscriptionConfig object or as keyword arguments:

from soniox import SonioxClient
from soniox.languages import Language
from soniox.types import TranscriptionConfig

client = SonioxClient()

# Using TranscriptionConfig
config = TranscriptionConfig(
    model="stt-async-preview",
    language_hints=[Language.en],
    enable_speaker_diarization=True,
    context="Medical terminology context"
)
job = client.transcribe_file("audio.wav", config=config)

# Or using kwargs
job = client.transcribe_file(
    "audio.wav",
    model="stt-async-preview",
    enable_speaker_diarization=True
)

Advanced Features

Speaker Diarization

Identify different speakers in your audio:

import time
from soniox import SonioxClient

client = SonioxClient()

# Submit job with speaker diarization
job = client.transcribe_file(
    "path/to/audio.wav",
    enable_speaker_diarization=True
)

# Wait for completion
while True:
    job = client.get_transcription_job(job.id)
    if job.status == "completed":
        break
    time.sleep(1)

# Get results with speaker information
result = client.get_transcription_result(job.id)
for token in result.tokens:
    if token.speaker:
        print(f"Speaker {token.speaker}: {token.text}")

Language Identification

Automatically identify the language being spoken:

from soniox import SonioxClient
from soniox.languages import Language

client = SonioxClient()

job = client.transcribe_file(
    "multilingual_audio.wav",
    language_hints=[Language.en, Language.es, Language.fr],
    enable_language_identification=True
)

Context for Improved Accuracy

Provide context to improve recognition of domain-specific terms:

from soniox import SonioxClient

client = SonioxClient()

job = client.transcribe_file(
    "medical_audio.wav",
    context="Medical terminology: hypertension, cardiovascular, stethoscope"
)

Configuration

Client Options

from soniox import SonioxClient

client = SonioxClient(
    api_key="your-api-key",           # API key (or use SONIOX_API_KEY env var)
    base_url="https://api.soniox.com", # Custom base URL (optional)
    timeout=60.0                       # Request timeout in seconds
)

Logging

The SDK uses Python's standard logging module with the logger name soniox:

import logging

# Enable debug logging
logging.basicConfig(level=logging.DEBUG)
logger = logging.getLogger("soniox")
logger.setLevel(logging.DEBUG)

# Or configure it your way
import logging

handler = logging.StreamHandler()
handler.setLevel(logging.INFO)
formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
handler.setFormatter(formatter)

logger = logging.getLogger("soniox")
logger.addHandler(handler)
logger.setLevel(logging.INFO)

API Reference

SonioxClient

Main client for interacting with Soniox API.

Methods

transcribe_file(file_path, config=None, **kwargs) → TranscriptionJob

Submit an audio file for transcription.

Parameters:

• file_path (str): Path to audio file
• config (TranscriptionConfig, optional): Configuration object
• **kwargs: Configuration options (used if config is None)
  • model (str): Model to use (default: "stt-async-preview")
  • language_hints (list[Language]): Language hints for better accuracy
  • enable_speaker_diarization (bool): Enable speaker diarization
  • enable_language_identification (bool): Enable language identification
  • context (str): Context for improved accuracy
  • webhook_url (str): Webhook URL for completion notification
  • client_reference_id (str): Your reference ID

Returns: TranscriptionJob - Job object with status information

Raises:

• FileNotFoundError: If file doesn't exist
• SonioxAPIError: If API returns an error

get_transcription_job(job_id) → TranscriptionJob

Get the status of a transcription job.

Parameters:

• job_id (str): Job ID from transcribe_file()

Returns: TranscriptionJob - Updated job status

get_transcription_result(job_id) → TranscriptionResult

Get the transcript once the job is completed.

Parameters:

• job_id (str): Job ID from completed transcription

Returns: TranscriptionResult - Transcript with tokens

Raises:

• SonioxAPIError: If job is not completed or not found

transcribe_file_async(file_path, config=None, **kwargs) → TranscriptionJob

Async version of transcribe_file().

get_transcription_job_async(job_id) → TranscriptionJob

Async version of get_transcription_job().

get_transcription_result_async(job_id) → TranscriptionResult

Async version of get_transcription_result().

Models

TranscriptionJob

Transcription job status and metadata.

Fields:

• id (str): Job ID (UUID)
• status (TranscriptionJobStatus): Job status ("queued", "processing", "completed", "error")
• created_at (datetime): Job creation timestamp
• filename (str): Original filename
• file_id (str | None): Uploaded file ID
• audio_url (str | None): Audio URL if provided
• audio_duration_ms (int | None): Audio duration in milliseconds
• error_message (str | None): Error message if failed
• All configuration fields from TranscriptionConfig

TranscriptionResult

Transcription result with full transcript.

Fields:

• id (str): Transcript ID (matches job ID)
• text (str): Full transcribed text
• tokens (list[Token]): Word-level tokens with timing

Token

Word-level transcription token.

Fields:

• text (str): Token text
• start_ms (int): Start time in milliseconds
• end_ms (int): End time in milliseconds
• confidence (float): Confidence score (0-1)
• speaker (str | None): Speaker ID if diarization enabled

TranscriptionConfig

Configuration for transcription jobs.

Fields:

• model (str): Model to use (default: "stt-async-preview")
• language_hints (list[Language] | None): Language hints
• enable_language_identification (bool): Enable language detection
• enable_speaker_diarization (bool): Enable speaker diarization
• context (str | None): Context for improved accuracy
• client_reference_id (str | None): Your reference ID
• webhook_url (str | None): Webhook URL
• webhook_auth_header_name (str | None): Webhook auth header name
• webhook_auth_header_value (str | None): Webhook auth header value

FileUploadResponse

Response from file upload.

Fields:

• id (str): File ID
• filename (str): Original filename
• size (int): File size in bytes
• created_at (datetime): Upload timestamp
• client_reference_id (str | None): Your reference ID

Exceptions

• SonioxError: Base exception for all Soniox errors
• SonioxAuthenticationError: Raised when authentication fails
• SonioxAPIError: Raised when API returns an error response
• SonioxRateLimitError: Raised when rate limit is exceeded

Error Handling

import time
from soniox import SonioxClient
from soniox.exceptions import (
    SonioxAPIError,
    SonioxAuthenticationError,
    SonioxRateLimitError,
)

client = SonioxClient()

try:
    # Submit transcription
    job = client.transcribe_file("audio.wav")
    
    # Wait for completion
    while True:
        job = client.get_transcription_job(job.id)
        if job.status == "completed":
            break
        elif job.status == "error":
            print(f"Transcription failed: {job.error_message}")
            break
        time.sleep(1)
    
    # Get result
    if job.status == "completed":
        result = client.get_transcription_result(job.id)
        print(result.text)

except FileNotFoundError:
    print("Audio file not found")
except SonioxAuthenticationError as e:
    print(f"Authentication failed: {e}")
except SonioxRateLimitError as e:
    print(f"Rate limit exceeded: {e}")
    print(f"Status code: {e.status_code}")
except SonioxAPIError as e:
    print(f"API error: {e}")
    print(f"Status code: {e.status_code}")
    print(f"Response: {e.response_body}")

Testing

Run tests with pytest:

# Install development dependencies
pip install -e ".[dev,test]"

# Run all tests
pytest

# Run with coverage
pytest --cov=src --cov-report=html --cov-report=term-missing

# Run specific test file
pytest tests/test_models.py

# Run with verbose output
pytest -v

See tests/README.md for more details on the test suite.

Development

# Clone the repository
git clone https://github.com/mahdikiani/soniox-sdk.git
cd soniox

# Install in editable mode with dev dependencies
pip install -e ".[dev,test]"

# Run linter
ruff check src/

# Run type checker
mypy src/

Resources

• Soniox Documentation
• API Reference
• GitHub Repository
• Issue Tracker

License

This project is licensed under the MIT License - see the LICENSE.txt file for details.

Support

• 📧 Email: mahdikiany@gmail.com
• 🐛 Issues: GitHub Issues
• 💬 Discussions: GitHub Discussions

Changelog

See CHANGELOG.md for version history and updates.

---

Made with ❤️ by Mahdi Kiani