Skip to main content
The official Python SDK is published as supertone on PyPI. Source: supertone-inc/supertone-python.

​
At a glance

Packagesupertone on PyPI
Reposupertone-inc/supertone-python
LanguagesPython 3.9+
AuthSupertone(api_key=...)
Sync APIYes (default)
Async API*_async methods + async with
Streamingiter_bytes() / aiter_bytes()
Auto-chunk long text✅ 300 chars, parallel up to 3 workers
Custom retries✅ via retry_config
HTTP backendhttpx

​
Installation

pip install supertone
Requires Python 3.9+.

​
Set your API key

The SDK does not auto-read from the environment. Pass api_key explicitly — the convention is to read from SUPERTONE_API_KEY. 
export SUPERTONE_API_KEY="Kp9mZ3xQ7v..."

import os
from supertone import Supertone

client = Supertone(api_key=os.environ["SUPERTONE_API_KEY"])
Voice IDs are not environment variables — they change per use case, so keep them as plain strings in your code (or pass them from your request payload).

​
Generate speech (sync)

The recommended pattern uses a context manager so the underlying HTTP connection is closed cleanly: 
import os
from supertone import Supertone

VOICE_ID = "20160a4c5ba38967330c84"  # replace with your voice ID

with Supertone(api_key=os.environ["SUPERTONE_API_KEY"]) as client:
    response = client.text_to_speech.create_speech(
        voice_id=VOICE_ID,
        text="Hello from the Python SDK.",
        language="en",
        output_format="wav",
    )

    with open("speech.wav", "wb") as f:
        f.write(response.result.read())

​
Generate speech (async)

Use the _async suffix and async with: 
import asyncio
import os
from supertone import Supertone

VOICE_ID = "20160a4c5ba38967330c84"  # replace with your voice ID

async def main():
    async with Supertone(api_key=os.environ["SUPERTONE_API_KEY"]) as client:
        response = await client.text_to_speech.create_speech_async(
            voice_id=VOICE_ID,
            text="Hello from the async Python SDK.",
            language="en",
        )

        with open("speech.wav", "wb") as f:
            f.write(response.result.read())

asyncio.run(main())
Every resource method on the SDK has both forms: create_speech / create_speech_async, stream_speech / stream_speech_async, list_voices / list_voices_async, and so on.

​
Stream speech

Streaming returns an iterator (or async iterator) of audio chunks: 
import os
from supertone import Supertone

VOICE_ID = "20160a4c5ba38967330c84"  # replace with your voice ID

with Supertone(api_key=os.environ["SUPERTONE_API_KEY"]) as client:
    response = client.text_to_speech.stream_speech(
        voice_id=VOICE_ID,
        text="This response is streamed chunk by chunk.",
        language="en",
        model="sona_speech_1",
    )

    with open("streamed.wav", "wb") as f:
        for chunk in response.result.iter_bytes():
            f.write(chunk)
Async equivalent uses async for chunk in response.result.aiter_bytes(). Streaming is currently supported on sona_speech_1 only.

​
Long text auto-chunking

create_speech, create_speech_async, stream_speech, and stream_speech_async automatically split text longer than 300 characters. create_speech runs up to 3 segments in parallel and merges the audio; stream_speech runs segments sequentially and forwards chunks to your iterator. 
LONG_TEXT = "..."  # any length, including thousands of characters

response = client.text_to_speech.create_speech(
    voice_id=VOICE_ID,
    text=LONG_TEXT,
    language="en",
)

with open("narration.wav", "wb") as f:
    f.write(response.result.read())   # single merged file
predict_duration does not auto-chunk — keep that input under 300 characters and sum durations manually for longer scripts. See Long text for details and tuning.

​
Common operations

# List voices with pagination
result = client.voices.list_voices(page_size=20)

# Search voices
result = client.voices.search_voices(language="ko,en", style="happy")

# Get a single voice
voice = client.voices.get_voice(voice_id=VOICE_ID)

# Predict duration (no credits deducted)
duration = client.text_to_speech.predict_duration(
    voice_id=VOICE_ID,
    text="How long will this be?",
    language="en",
)

# Get credit balance
balance = client.usage.get_credit_balance()

​
Type-safe enums (optional)

For type safety, the SDK exposes enum constants in supertone.models: 
from supertone import models

models.APIConvertTextToSpeechUsingCharacterRequestLanguage.EN       # "en"
models.APIConvertTextToSpeechUsingCharacterRequestModel.SONA_SPEECH_1  # "sona_speech_1"
Plain strings ("en", "sona_speech_1") work too — use whichever style you prefer.

​
Error handling

Errors live in supertone.errors and all extend SupertoneError: 
from supertone import Supertone, errors

try:
    response = client.text_to_speech.create_speech(...)
except errors.TooManyRequestsErrorResponse as e:
    # 429 — back off and retry
    print("Rate limited:", e.message)
except errors.UnauthorizedErrorResponse as e:
    # 401 — bad or missing API key
    print("Auth failed:", e.message)
except errors.PaymentRequiredErrorResponse as e:
    # 402 — out of credits
    print("Buy more credits:", e.message)
except errors.SupertoneError as e:
    # Any other API error
    print(f"HTTP {e.status_code}: {e.message}")
Error classHTTP status
BadRequestErrorResponse400
UnauthorizedErrorResponse401
PaymentRequiredErrorResponse402
ForbiddenErrorResponse403
NotFoundErrorResponse404
RequestTimeoutErrorResponse408
PayloadTooLargeErrorResponse413
UnsupportedMediaTypeErrorResponse415
TooManyRequestsErrorResponse429
InternalServerErrorResponse500
Network errors (DNS failure, broken pipe, etc.) come from httpx and don’t inherit from SupertoneError.

​
Configuration

from supertone import Supertone
from supertone.utils.retries import RetryConfig

client = Supertone(
    api_key=os.environ["SUPERTONE_API_KEY"],
    timeout_ms=30_000,
    retry_config=RetryConfig(
        strategy="backoff",
        backoff={"initial_interval": 500, "max_interval": 60_000},
        retry_connection_errors=True,
    ),
)

TypeScript SDK

The equivalent SDK for Node and Bun.

Examples

Recipes for common workflows.