whisperpipe 0.1.1

Real-time speech-to-text streaming with OpenAI Whisper

whisperpipe

Real-time speech-to-text streaming with OpenAI Whisper

Description

whisperpipe is a powerful, easy-to-use Python package for real-time, offline audio transcription using OpenAI's Whisper model. It runs locally, making it a free and private solution for continuous speech-to-text applications. It provides seamless integration with callback functions for LLM processing and supports pause/resume functionality for interactive applications.

Why whisperpipe?

In a world where most ASR (Automatic Speech Recognition) services are cloud-based, whisperpipe offers a refreshing alternative by harnessing the power of OpenAI's Whisper model to run directly on your local machine. This approach provides several key advantages:

• Complete Privacy: Since all transcription is done locally, your voice data never leaves your computer. This is crucial for applications that handle sensitive or private conversations.
• Zero Cost: Say goodbye to recurring subscription fees and per-minute charges. whisperpipe is free to use, making it an economical choice for both hobbyists and commercial projects.
• No Internet Required: Whether you're on a plane, in a remote location, or simply have an unstable internet connection, whisperpipe works flawlessly offline.
• Real-time Performance: Designed for continuous, real-time transcription, whisperpipe is ideal for live applications such as voice-controlled assistants, dictation software, and more.
• Unleash the Power of Whisper: By running the Whisper model locally, you have full control over the transcription process, from model selection to performance tuning.

whisperpipe empowers you to build powerful, private, and cost-effective voice applications with ease.

Features

• Real-time audio transcription using OpenAI Whisper
• Callback system for custom processing (LLM integration, etc.)
• Pause/Resume functionality for interactive applications
• Multiple language support
• Configurable processing parameters
• Thread-safe operation
• Easy installation and usage

Installation

From PyPI

pip install whisperpipe

From GitHub

pip install git+https://github.com/Erfan-ram/whisperpipe.git

Quick Start

from whisperpipe import pipeStream

# Basic usage
transcriber = pipeStream(
    model_name="base",
    language="en",
    finalization_delay=10.0,
    processing_interval=1.0
)

# Start streaming
transcriber.start_streaming()

Usage Examples

Basic Transcription

from whisperpipe import pipeStream

# Create transcriber instance
transcriber = pipeStream(
    model_name="base",
    language="en",
    finalization_delay=10.0,
    processing_interval=1.0
)

# Start transcription
transcriber.start_streaming()

# The transcribed text will be printed to console
# Press Ctrl+C to stop

With Custom Callback (LLM Integration)

from whisperpipe import pipeStream

def llm_processor(text):
    """Custom function to process transcribed text"""
    print(f"Processing: {text}")
    # Your LLM integration here
    # e.g., send to OpenAI, Claude, local model, etc.
    response = your_llm_api.chat(text)
    print(f"Response: {response}")
    return response

# Create transcriber with callback
transcriber = pipeStream(
    model_name="base",
    language="en",
    finalization_delay=10.0,
    processing_interval=1.0
)

# Register callback
transcriber.set_def_callback(llm_processor)

# Start streaming with LLM integration
transcriber.start_streaming()

Interactive Mode with Pause/Resume

from whisperpipe import pipeStream
import time

def interactive_processor(text):
    """Process text and pause for response"""
    # Pause transcriber while processing
    transcriber.pause_streaming()
    
    print(f"User said: {text}")
    
    # Process with your system
    response = process_with_llm(text)
    
    # Speak or display response
    print(f"Assistant: {response}")
    
    # Resume for next input
    transcriber.resume_streaming()

transcriber = pipeStream()
transcriber.set_def_callback(interactive_processor)
transcriber.start_streaming()

API Reference

Constructor Parameters

• model_name (str): Whisper model name ("tiny", "base", "small", "medium", "large"). Default: "base"
• language (str): Language code for transcription ("en", "es", "fr", etc.). Default: "en"
• finalization_delay (float): Wait time in seconds before finalizing transcription. Default: 10.0
• processing_interval (float): Interval in seconds between processing cycles. Default: 1.0
• buffer_duration_seconds (float): Time window in seconds to hold audio for processing. Default: 5.0
• debug_mode (bool): Enable debug mode for detailed logging. Default: True

Methods

Core Methods

• start_streaming(): Start audio capture and transcription
• stop_streaming(): Stop audio capture and transcription

Callback System

• set_def_callback(callback_function): Register a callback function for processing transcribed text
• set_def_callback(None): Clear the callback (use default behavior)

Pause/Resume Control

• pause_streaming(): Pause audio processing temporarily
• resume_streaming(): Resume audio processing
• is_paused(): Check if transcriber is paused
• is_running(): Check if transcriber is running

Requirements

• Python 3.8+
• PyAudio
• OpenAI Whisper
• PyTorch
• NumPy
• pynput

License

MIT License

Author

Erfan Ramezani - [e-mail]

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Support

For issues and questions, please use the GitHub Issues page.

Citation

If you use whisperpipe in your research, please cite both our paper and the software repository.

Paper (arXiv):

The arXiv preprint link will be available here shortly. Once published, the formal BibTeX will be updated.

Software / Codebase (Zenodo):

**Software / Codebase (Zenodo):**
```bibtex
@software{whisperpipe_code_2026,
  author       = {Erfan Ramezani and Mohammad Mahdi Giahi},
  title        = {WhisperPipe: Source Code and Implementation},
  month        = apr,
  year         = 2026,
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.19646625},
  url          = {[https://doi.org/10.5281/zenodo.19646625](https://doi.org/10.5281/zenodo.19646625)}
}