Gradium

The GradiumTTSService provides high-fidelity, low-latency text-to-speech using Gradium's specialized neural voice technology. It features a persistent WebSocket connection for sub-second synthesis and sub-word timestamp support.

Installation

To use Gradium, install the required dependencies:

pip install "piopiy-ai[gradium]"

Prerequisites

• A Gradium account and API key (Get yours here).
• Set your API key in your environment:

  export GRADIUM_API_KEY="your_api_key_here"

Configuration

GradiumTTSService Parameters

Parameter	Type	Default	Description
api_key	str	Required	Your Gradium API key.
voice_id	str	"YTpq7expH9539ERJ"	The voice identifier to use.
url	str	wss://eu.api.gradium.ai/api/speech/tts	Gradium WebSocket endpoint.
model	str	"default"	Model ID to use for synthesis.
json_config	str	None	Optional JSON for additional model settings.
params	InputParams	InputParams()	Advanced generation settings.

InputParams

Parameter	Type	Default	Description
temp	float	0.6	Temperature for generation randomness.

Usage

Basic Setup

import os
from piopiy.services.gradium.tts import GradiumTTSService

tts = GradiumTTSService(
    api_key=os.getenv("GRADIUM_API_KEY"),
    voice_id="YTpq7expH9539ERJ"
)

Notes

• WebSocket Connection: Gradium uses a persistent WebSocket connection to minimize handshaking latency during live calls.
• High Fidelity: Default output is mono PCM at 48kHz, providing rich and articulate voice quality.
• Word Timestamps: Supports real-time word-level timing information, enabling synchronized visual cues and lip-syncing.