Pydub (Pozalabs Fork)

Note: This is a fork tailored for Pozalabs internal use. The API may differ significantly from upstream.

Pydub lets you do stuff to audio in a way that isn't stupid.

This is a Pozalabs fork of jiaaro/pydub, published as pozalabs-pydub on PyPI.

Changes from Upstream

Requirements

• Python 3.11+ (upstream supports Python 2.7+)
• All legacy Python 2 compatibility code has been removed

Performance (Rust/PyO3 Extensions)

• overlay_segments - Pre-allocated buffer overlay replacing audioop.add/mul (16-bit 4-5x, 32-bit 11-16x faster)
• extend_24bit_to_32bit - Zero-copy 24-bit to 32-bit sample extension via direct PyBytes allocation (~400x faster than pure Python)
• fade_segment - Single-pass fade replacing per-sample audioop.mul loop (used by fade/fade_in/fade_out). Upstream applies gain per millisecond (>100ms) or per sample (<=100ms); this implementation always applies gain per sample for consistent precision
• mix_segments - N-segment single-pass mix in Rust (~2.9x faster than sequential overlay chaining with 19 segments)

New Features

• Compressed audio I/O - from_file() auto-detects and decompresses gzip/zstd files; export() accepts an optional compressor parameter
• Audio level metering - measure_audio_level() for RMS, peak, and LUFS measurement
• Waveform data - get_normalized_amplitudes() computes normalized amplitude values for waveform visualization
• Silent audio generation - Create silent audio matching the original segment's parameters
• Multi-segment mix - AudioSegment.mix(*segs) classmethod for mixing N segments simultaneously
• Audio processing framework - Command-based processor architecture for merge, overlay, and format conversion
• Python 3.13 support - Via audioop-lts dependency

Removed

• from_file_using_temporary_files - Redundant legacy method replaced by from_file (pipe I/O, codec inference, compression support)
• AudioSegment.ffmpeg - Legacy class property alias for AudioSegment.converter
• effects / scipy_effects modules - All dynamically registered effects (normalize, speedup, compress_dynamic_range, invert_phase, low_pass_filter, high_pass_filter, pan, apply_gain_stereo, strip_silence, eq, etc.) and the register_pydub_effect decorator

Type Safety

• Comprehensive type hints using Self, Literal, TypedDict, Unpack
• _AudioParams dataclass with validation for initialization parameters

Build System

• pyproject.toml with maturin backend, managed by uv
• Rust(PyO3) native extension compiled via maturin
• Wheel distribution via GitHub Actions

Dependencies

• backports.zstd (>=1.3.0) - Zstandard compression (Python <3.14, stdlib compression.zstd on 3.14+)

For general usage, API documentation, and ffmpeg setup, see the upstream README.

License

MIT License - Copyright 2011 James Robert