voice-of-python 0.1.1

Offline Multilingual Voice Bot Library for Python

Voice of Python

Offline Multilingual Voice Bot Library for Python 🎙️

voice_of_python is a robust Python library providing offline multilingual speech recognition and text-to-speech (TTS) capabilities. Powered by the Vosk toolkit for precise offline speech-to-text and pyttsx3 for offline TTS, it seamlessly supports multiple languages with automatic language detection.

---

🌟 Key Features

• Offline Processing: Complete offline speech-to-text using Vosk models—no APIs or internet required.
• Audio Inputs: Record audio directly from the microphone or process existing WAV files.
• Auto-Language Detection: Smartly identifies spoken languages using langid.
• Text-to-Speech (TTS): Synthesizes responses in the detected language's voice (if locally available).
• CPU Optimized: Fully functional on CPU, making it perfect for local, privacy-centric applications.
• LLM Ready: Purpose-built for easy integration with large language models (LLMs) and custom chatbots.

---

📦 Installation

Ensure you have installed the required dependencies. You can install the package directly from PyPI:

pip install voice_of_python

(Alternatively, to install dependencies manually: pip install vosk langid pyttsx3 sounddevice numpy soundfile)

Note: Download and extract Vosk models for your supported languages separately to a local folder.

---

🚀 Quick Start

1. Initialize the Voice Bot

Provide the paths to your downloaded Vosk models:

from voice_of_python import MultiLingualVoiceBot

model_paths = {
    "en": "/path/to/vosk-model-en-us-0.22",
    "hi": "/path/to/vosk-model-small-hi-0.22"
}

bot = MultiLingualVoiceBot(model_paths)

2. Record & Transcribe Live Audio

Listen from the microphone, detect language, and transcribe text:

text, lang_code = bot.record_and_transcribe(record_duration=5)

print(f"Detected language: {lang_code}")
print(f"Recognized text: {text}")

3. Integrate with an LLM (Example)

Pass the transcribed text to your backend model (like OpenAI, LLaMA, etc.) to generate a conversational response.

# Example response from an LLM
reply_text = "यहाँ आपका उत्तर है"

4. Speak the Reply

Respond to the user naturally in the detected language:

bot.speak_text(reply_text, lang_code)

5. Process Existing Audio Files

Alternatively, handle pre-recorded audio files:

audio_file = "user.wav"
reply_text = "Thank you for your question."

bot.process_audio_and_reply(audio_file, reply_text)

---

📖 API Reference

MultiLingualVoiceBot(model_paths: dict, sample_rate: int = 16000)

Initializes the bot.

• model_paths: A dictionary mapping language codes to their respective local Vosk model paths (e.g., {"en": "/path"}).
• sample_rate: Sampling rate for audio processing (default is 16000 Hz).

record_and_transcribe(record_duration: int = 5) -> tuple[str, str]

Records audio directly from the default microphone.

• Returns: A tuple containing (recognized_text, language_code).

process_audio_and_reply(audio_input: Union[str, np.ndarray], reply_text: str)

Transcribes provided audio, detects language, and immediately speaks the provided reply.

• audio_input: A file path to a WAV file or a raw numpy array.
• reply_text: The text response to be spoken aloud.

speak_text(text: str, lang_code: str)

Reads out the text using the system's text-to-speech engine.

• text: Text to synthesize.
• lang_code: The language code to select the proper voice.

---

🤝 Contributing

We welcome contributions! Please feel free to open issues, submit pull requests, or discuss new features on GitHub.

---

📚 Acknowledgements

This project builds upon several amazing open-source libraries:

• Vosk: For robust offline speech recognition.
• pyttsx3: For text-to-speech conversion.
• langid: For accurate language identification.

---

💬 Support

If you encounter any bugs or have questions, please open an issue on GitHub.

Enjoy building multilingual, offline voice-enabled applications with voice_of_python!