当前版本 Version: 0.2.0

🆕 0.2.0 更新日志 / Changelog:

• 修复了上一版本的初始延迟导致的结果偏移的问题。
• Fixed the initial latency/offset issue in previous versions.
• 建议所有用户升级到 0.2.0 版本。

该项目提供了 Bungee C++ 库的 Python 绑定，允许您在 Python 中轻松实现高质量的实时音频时间拉伸和变调。

This project provides Python bindings for the Bungee C++ library, enabling high-quality real-time audio time-stretching and pitch-shifting in Python.

• 高质量的音频时间拉伸和变调
  High-quality audio time-stretching and pitch-shifting
• 支持实时处理，低延迟
  Real-time processing with low latency
• 允许连续改变播放速度和音高，支持平滑搓碟和变速播放
  Continuous speed/pitch change, smooth scratching, and variable playback
• 基于频域相位声码器算法
  Frequency-domain phase vocoder algorithm
• 使用现代 C++ 编写，代码清晰健壮
  Modern C++ implementation, robust and clean code
• 采用 MPL-2.0 宽松许可证
  Licensed under MPL-2.0

您可以通过 pip 从 PyPI 安装 bungee-python:

You can install bungee-python from PyPI via pip:

pip install bungee-python

下面是一个简单的示例，演示如何使用 bungee-python 将音频速度减慢一半：

Here is a simple example showing how to slow down audio by half using bungee-python:

import numpy as np from bungee_python import bungee sample_rate = 44100 channels = 1 duration_seconds = 5 frequency = 440 t = np.linspace(0., duration_seconds, int(sample_rate * duration_seconds)) input_audio = 0.5 * np.sin(2. * np.pi * frequency * t) input_audio = input_audio.astype(np.float32) input_audio = input_audio[:, np.newaxis] # (frames, channels) stretcher = bungee.Bungee(sample_rate=sample_rate, channels=channels) stretcher.set_speed(0.5) # 慢放 slow down stretcher.set_pitch(1.0) # 音高不变 keep pitch output_audio = stretcher.process(input_audio) print(f"Output shape: {output_audio.shape}")

1. 克隆仓库 (包括子模块) / Clone repository (with submodules):

   git clone --recurse-submodules https://github.com/longredzhong/bungee-python.git cd bungee-python

2. 安装构建依赖 / Install build dependencies:

   • C++ 编译器 (支持 C++17) / C++17 compiler
   • CMake (>= 3.15)
   • Ninja (推荐 recommended)
   • Python (>= 3.8) 和开发头文件 / Python dev headers

3. 运行构建脚本 / Run build script:

   ./scripts/build.sh

   编译后的 Python 扩展模块会位于 build 目录中。
   The built Python extension module will be in the build directory.

4. 安装或测试 / Install or test:

   可以用 pip install . 在本地安装。
   You can install locally with pip install ..

• 运行时 Runtime:
  • Python (>= 3.8)
  • NumPy
• 构建时 Build:
  • bungee-core (as submodule)
    • Eigen
    • KissFFT
  • pybind11 (as submodule)
  • CMake (>= 3.15)
  • C++17 compiler
  • Ninja (optional)

本项目采用 MPL-2.0 许可证，与 bungee-core 保持一致。
This project is licensed under MPL-2.0, same as bungee-core.

• 感谢 Parabola Research Limited 开发了优秀的 Bungee 库。
  Thanks to Parabola Research Limited for developing the excellent Bungee library.
• 感谢 pybind11 团队提供了方便易用的 C++/Python 绑定工具。
  Thanks to the pybind11 team for their great C++/Python binding tool.