OuteAI API & SDK

Introduction

OuteAI provides high-quality text-to-speech generation through two interfaces:

• 1 Studio — a browser-based interface for generating and managing audio. No code required. Great for one-off generations, content production, and voice experimentation.
• 2 API / SDK — a REST API with an official Python SDK for building applications, automating workflows, or integrating TTS into your product.

Both interfaces use the same underlying model, voice library, and credits balance. Audio is billed per second of generated audio at 0.001 credits / second.

GPU Cold Starts

This applies to both Studio and the API. Subsequent requests within the same active window are fast.

In practice: if you are building a latency-sensitive application, consider sending a short warm-up request (e.g., a single word) when your application starts, then making real requests immediately after. You will still be billed normally for the warm-up audio.

Authentication

All API requests are authenticated with a Bearer token. You can create and manage API tokens from your Account page.

Authorization: Bearer oute_xxxxxxxxxxxxxxxxxxxxxxxx

Using Studio

Studio is a browser-based interface at /studio. Sign in to get started — no setup required.

Voice Cloning

Create a custom voice from a reference audio sample in the Voice Clone section of Studio.

Audio requirements

Once cloned, your voice appears in Studio's voice selector and is available via the API using its UUID. Clones can be deleted at any time from the Voice Clone page.

Voice cloning is billed at 0.025 credits / voice. Clones are stored until you delete them.

Installation

Install the outeai package from PyPI. Python 3.9+ is required.

# Synchronous client only
pip install outeai

# Synchronous + async client (installs httpx)
pip install "outeai[async]"

Quick Start

List available voices, generate speech, and save it to a file.

from outeai import OuteAI

# Initialise the client — use a context manager to auto-close
with OuteAI("oute_xxxxxxxxxxxxxxxxxxxxxxxx") as client:

    # List all voices available on your account
    voices = client.list_voices()
    for v in voices:
        print(v["voice_id"], v["speaker_name"])

    # Generate speech (non-streaming, returns full WAV)
    audio = client.generate_speech(
        text="Hello, world! This is OuteAI.",
        voice_id=voices[0]["voice_id"],
    )

    # Save to a WAV file
    audio.save("hello.wav")

Generate and save in one call

client.generate_speech_to_file(
    "output.wav",
    text="Saving directly to a file.",
    voice_id="your-voice-uuid",
)

SpeechResult object

generate_speech() returns a SpeechResult with the following attributes:

Streaming

Use stream_speech() to receive audio chunks as they are generated. This is useful for real-time playback or piping audio to another process.

with client.stream_speech(
    text="Streaming audio chunk by chunk.",
    voice_id="your-voice-uuid",
) as stream:
    # Iterate over raw WAV chunks
    for chunk in stream:
        my_audio_pipe.write(chunk)

Stream directly to a file

with client.stream_speech(
    text="Saving a stream.",
    voice_id="your-voice-uuid",
) as stream:
    stream.save("streamed.wav")

# Or use the one-liner shorthand
client.stream_speech_to_file(
    "streamed.wav",
    text="Saving a stream.",
    voice_id="your-voice-uuid",
)

Async Client

Install with async support and use AsyncOuteAI for non-blocking generation in async applications.

pip install "outeai[async]"

import asyncio
from outeai import AsyncOuteAI

async def main():
    async with AsyncOuteAI("oute_xxxxxxxxxxxxxxxxxxxxxxxx") as client:
        voices = await client.list_voices()

        audio = await client.generate_speech(
            text="Hello from async!",
            voice_id=voices[0]["voice_id"],
        )
        audio.save("async-output.wav")

asyncio.run(main())

Async streaming

async with client.stream_speech(
    text="Async streaming example.",
    voice_id="your-voice-uuid",
) as stream:
    async for chunk in stream:
        my_pipe.write(chunk)

# Or save directly
await client.stream_speech_to_file(
    "streamed.wav",
    text="Async streaming to file.",
    voice_id="your-voice-uuid",
)

Error Handling

The SDK raises two exception types:

from outeai import OuteAI, OuteAIError, OuteAIAPIError

try:
    audio = client.generate_speech(text="Hello", voice_id="...")
except OuteAIAPIError as e:
    # HTTP error returned by the API
    print(e.status_code)  # e.g. 402, 429, 400
    print(e.code)         # e.g. "INSUFFICIENT_CREDITS"
    print(e.message)      # human-readable description
except OuteAIError as e:
    # Network error or SDK-level validation failure
    print(e)

List Voices

{
    "success": True,
    "data": {
        "voices": [
            {
                "voice_id": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
                "speaker_name": "My Cloned Voice",
                "is_clone": True,
            }
        ]
    },
}

Generate Speech

curl -X POST https://outeai.com/api/v1/tts_non_stream \
  -H "Authorization: Bearer oute_xxxx" \
  -H "Content-Type: application/json" \
  -H "Accept: audio/wav" \
  -d '{
    "text": "Hello, world!",
    "voice_id": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
    "model_name": "outetts_1_pro"
  }' \
  --output hello.wav

Streaming Speech

curl -X POST https://outeai.com/api/v1/tts_stream \
  -H "Authorization: Bearer oute_xxxx" \
  -H "Content-Type: application/json" \
  -H "Accept: audio/wav" \
  -d '{
    "text": "Streaming audio in real time.",
    "voice_id": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
    "model_name": "outetts_1_pro"
  }' \
  --output streamed.wav

Voice Clones

curl -X POST https://outeai.com/api/v1/voice-clones \
  -H "Authorization: Bearer oute_xxxx" \
  -F "speaker_name=My Voice" \
  -F "audio=@/path/to/sample.wav"

curl -X POST https://outeai.com/api/v1/voice-clones \
  -H "Authorization: Bearer oute_xxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "speaker_name": "My Voice",
    "audio_base64": "<base64-encoded-audio>"
  }'

API Tokens

Models

Additional models may be added in future. Always pass the model ID string in your requests.

Error Codes

All error responses share the same JSON envelope:

{
    "success": False,
    "code": "INSUFFICIENT_CREDITS",
    "message": "You do not have enough credits to start this request.",
}