voicepipe 0.1.0

One-command STT + TTS for any app

VoicePipe

One-command voice integration for any app

---

Overview

VoicePipe provides one-command STT (Speech-to-Text) + TTS (Text-to-Speech) for any application.

• STT: whisper.cpp - fastest local speech recognition
• TTS: KittenTTS - smallest neural TTS (15-80MB)

Installation

pip install voicepipe

Quick Start

from voicepipe import VoicePipeline

# Initialize (auto-downloads models)
voice = VoicePipeline()

# Speech to Text
text = voice.speech_to_text("audio.wav")
print(f"You said: {text}")

# Text to Speech
audio = voice.text_to_speech("Hello, world!")

Requirements

• Python 3.8+
• FFmpeg (for audio processing)

Install FFmpeg

macOS:

brew install ffmpeg

Linux:

sudo apt install ffmpeg

Windows:

choco install ffmpeg

Configuration

voice = VoicePipeline(
    stt_model="tiny",        # tiny, base, small
    tts_model="nano",        # nano, micro, mini
    tts_voice="Bella",       # 8 voices available
    tts_speed=1.0,          # 0.5 - 2.0
    language="en",           # or "auto"
    cache_dir="~/.voicepipe" # model cache
)

Available Voices

• Bella, Jasper, Luna, Bruno, Rosie, Hugo, Kiki, Leo

Models

STT (whisper.cpp)

Model	Size	RAM	Speed
tiny	75MB	~500MB	10x realtime
base	142MB	~1GB	5x realtime
small	466MB	~2GB	2x realtime

TTS (KittenTTS)

Model	Size	Quality
nano	15MB	Good
micro	40MB	Better
mini	80MB	Best

Use Cases

Chatbot with Voice

@app.post("/voice/chat")
async def voice_chat(audio: bytes):
    # Convert speech to text
    text = voice.speech_to_text_bytes(audio)
    
    # Get chatbot response
    response = await chatbot.chat(text)
    
    # Convert response to speech
    audio_response = voice.text_to_speech(response)
    
    return {"audio": audio_response}

Voice Assistant

async def run_assistant():
    while True:
        # Continuously listen and respond
        text = await voice.speech_to_text_async(microphone_stream)
        response = await assistant.respond(text)
        voice.text_to_speech(response, play=True)

API Reference

VoicePipeline

Method	Description
speech_to_text(audio_path)	Convert audio file to text
speech_to_text_bytes(audio_data)	Convert raw audio to text
text_to_speech(text)	Convert text to audio bytes
text_to_speech_file(text, path)	Convert text to audio file
list_voices()	Get available TTS voices
get_status()	Get pipeline status

Development

# Clone repository
git clone https://github.com/danlab-ai/voicepipe.git
cd voicepipe

# Install in development mode
pip install -e ".[dev]"

# Run tests
pytest

# Format code
black src/voicepipe
ruff check src/voicepipe

License

MIT License - see LICENSE

---

Built by DanLab