XTTS (Local)

The XTTSService integrates with a locally running Coqui XTTS streaming server. It is ideal for applications requiring local voice cloning and absolute data privacy.

Installation

1. Run the XTTS Docker container:

   docker run --gpus=all -e COQUI_TOS_AGREED=1 --rm -p 8000:80 ghcr.io/coqui-ai/xtts-streaming-server:latest-cuda121
2. Install the SDK:

   pip install piopiy-ai

Configuration

XTTSService Parameters

Parameter	Type	Default	Description
voice_id	str	Required	ID of the speaker to use.
base_url	str	Required	URL of the local XTTS server.
language	Language	Language.EN	Language for synthesis.

Usage

Basic Setup

import aiohttp
from piopiy.services.xtts.tts import XTTSService

async with aiohttp.ClientSession() as session:
    tts = XTTSService(
        voice_id="female_voice",
        base_url="http://localhost:8000",
        aiohttp_session=session
    )

Notes

• Voice Cloning: You can load different speakers by passing their IDs retrieved from the /studio_speakers endpoint.
• Performance: XTTS generates audio at 24kHz; the SDK automatically resamples it to your desired rate.
• Privacy: Since it runs locally in Docker, no audio data or text is sent to an external service.