Hume AI

The HumeTTSService provides emotion-aware voice generation creating deeply empathetic and expressive agent responses. Leveraging Hume's Octave models, it allows for natural-sounding speech guided by descriptive acting directions.

Installation

To use Hume AI, install the required dependencies:

pip install "piopiy-ai[hume]"

Prerequisites

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

  export HUME_API_KEY="your_api_key_here"

Configuration

HumeTTSService Parameters

Parameter	Type	Default	Description
api_key	str	Required	Your Hume API key.
voice_id	str	Required	The ID of the voice to use.
sample_rate	int	48000	Audio sample rate (Hume streams at 48kHz).
params	InputParams	InputParams()	Acting directions and speed settings.

InputParams

Parameter	Type	Default	Description
description	str	None	Natural-language acting directions.
speed	float	None	Speaking-rate multiplier (0.5 to 2.0).
trailing_silence	float	None	Seconds of silence to append at the end.

Usage

Basic Setup

import os
from piopiy.services.hume.tts import HumeTTSService

tts = HumeTTSService(
    api_key=os.getenv("HUME_API_KEY"),
    voice_id="default"
)

With Acting Directions

from piopiy.services.hume.tts import HumeTTSService

tts = HumeTTSService(
    voice_id="default",
    params=HumeTTSService.InputParams(
        description="Speak in a warm, comforting, and empathetic tone.",
        speed=0.9
    )
)

Notes

• Empathic Voice: Hume's TTS is uniquely designed to convey emotional nuance; use the description parameter to fine-tune the agent's persona.
• Word Timestamps: Supports millisecond-precision word-level timestamps for perfect audio-text synchronization.
• Buffering: Piopiy automatically buffers audio chunks from Hume to ensure glitch-free streaming over telephony networks.