Gladia

The GladiaSTTService provides ultra-fast, real-time transcription using Gladia's optimized speech models. It includes advanced features like automatic language detection, translation, and high-performance VAD.

Installation

To use Gladia, install the required dependencies:

pip install "piopiy-ai[gladia]"

Prerequisites

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

  export GLADIA_API_KEY="your_api_key_here"

Configuration

GladiaSTTService Parameters

Parameter	Type	Default	Description
api_key	str	Required	Your Gladia API key.
sample_rate	int	None	Audio sample rate in Hz.
model	str	"solaria-1"	Model to use (e.g., "solaria-1").
params	GladiaInputParams	None	Advanced VAD and language settings.

GladiaInputParams

Parameter	Type	Default	Description
language_config	LanguageConfig	None	Multi-language and code-switching setup.
endpointing	int	None	Silence duration (ms) for segmenting.
enable_vad	bool	False	Enable Gladia's voice activity detection.

Usage

Basic Setup

import os
from piopiy.services.gladia.stt import GladiaSTTService
from piopiy.services.gladia.config import GladiaInputParams

stt = GladiaSTTService(
    api_key=os.getenv("GLADIA_API_KEY"),
    params=GladiaInputParams(
        endpointing=300,
        enable_vad=True
    )
)

Notes

• VAD Interruption: When enable_vad is true, Gladia will signal when a user starts and stops speaking, which can be used to interrupt the bot immediately.
• Language Detection: Gladia excelled at detecting language changes and code-switching in real-time.