VapourSynth Ares

This package contains the VS-AudioResample VapourSynth plugin.

Installation

pip install vapoursynth-ares

Building from source

uv build --package vapoursynth-ares

Detailed parameter information from the parent project follows.

VS-AudioResample

This is an audio sample rate and sample type converter for VapourSynth utilizing the SoX Resampler library.

Usage

ares.Resample(clip: vs.AudioNode,
              sample_rate: int = -1,
              sample_type: str = None,
              quality: str = 'very_high',
              overflow: str = 'error',
              overflow_log: str = 'once'
              ) -> vs.AudioNode

clip - input audio clip (any format)

sample_rate - new sample rate (e.g. 16000, 44100, 48000, ...); same as input clip if negative; default: -1

sample_type - new sample type; same as input clip if None; default: None

    'i16' - integer 16-bit
    'i24' - integer 24-bit
    'i32' - integer 32-bit
    'f32' - float   32-bit

quality - resample quality; default: 'very_high'

    'quick'     - quick cubic interpolation
    'low'       - low quality
    'medium'    - medium quality
    'high'      - high quality
    'very_high' - very high quality (default)
    'max'       - maximum quality

overflow - sample overflow handling; default: 'error'

    'error'      - raise an error (default)
    'clip'       - clip overflowing samples (all types)
    'clip_int'   - clip overflowing samples for integer output sample types
                   keep overflowing samples for float output sample types
    'keep_float' - keep overflowing samples for float output sample types
                   raise an error if output sample type is not float

To properly handle overflows the clip should be converted to a float sample type ('f32'), if not already.

⚠️ Overflowing samples of integer sample types (output) are always clipped (disruptive), or they raise an error

Use overflow='keep_float' for float output sample types to leave overflowing samples unchanged.
Then call a scaling function like std.AudioGain that scales the peak sample value below or to equal 1.0 (see Example 3)

overflow_log - sample overflow logging; default: 'once'

    'all'  - log all sample overflows (not recommended, this can be a lot)
    'once' - log only the first sample overflow (default)
    'none' - do not log any sample overflows

Note: a summary of all overflowing samples will be logged at the end of each function (if any)

Example 1

basic usage

import vapoursynth as vs

# load audio
audio = ...

# convert sample rate to 48000 and sample type to 24-bit integer
audio = vs.core.ares.Resample(audio, sample_rate=48000, sample_type='i24')

Example 2

change speed of an audio clip (but keep the sample rate)

import vapoursynth as vs

# load audio
audio = ...

# speedup factor
factor = 2.0

resample_rate = round(audio.sample_rate / factor)
res_audio = vs.core.ares.Resample(audio, sample_rate=resample_rate)
audio = vs.core.std.AssumeSampleRate(res_audio, samplerate=audio.sample_rate)

Example 3

handle overflowing samples

import vapoursynth as vs

# load audio (integer or float sample type)
audio = ...

# convert sample rate to 48000 and sample type to 32-bit float
# leave possible overflowing samples unchanged with 'keep_float'
audio = vs.core.ares.Resample(audio, sample_rate=48000, sample_type='f32', overflow='keep_float')

# scale audio samples
# choose a factor that limits the peak value below or to equal 1
audio = vs.core.std.AudioGain(audio, 0.5)

# optional: convert sample type back to integer if needed (e.g. 24-bit)
audio = vs.core.ares.Resample(audio, sample_type='i24')

Dependencies

This project uses the SoX Resampler library (soxr), which is licensed under the GNU Lesser General Public License (LGPL) v2.1.

The distributed binaries of this plugin statically link to soxr for performance and ease of use. If you prefer to use a customized version of soxr, or to link dynamically instead, you can build the plugin from source. See the Build from source section for instructions.

Build from source

To build the plugin, you’ll need CMake and a C++20-compatible compiler. OpenMP support is optional.

Run CMake to configure your preferred build system, then build the project. This process will automatically download and build the soxr dependency along with the plugin.

Note: You don’t need to download or build soxr yourself. However, if you prefer to use a local or customized version of soxr, you can provide its source path during configuration (see options below).

e.g. CMake with Ninja:

# EITHER build with statically linked soxr
cmake -G Ninja -B ./build-ninja -DBUILD_SHARED_LIBS=OFF -DCMAKE_BUILD_TYPE=Release

# OR build with dynamically linked soxr
cmake -G Ninja -B ./build-ninja -DBUILD_SHARED_LIBS=ON -DCMAKE_BUILD_TYPE=Release

ninja -C ./build-ninja

Custom options for cmake:

-DSOXR_SOURCE_DIR=<PATH>     Use a custom soxr source directory
                             (must contain CMakeLists.txt)
-DSOXR_USE_PATCHES=<ON|OFF>  Enable patches needed for outdated 
                             soxr build scripts (default: ON)
-DWITH_OPENMP=<ON|OFF>       Enable or disable OpenMP support (default: ON)

License

This project is licensed under the MIT License.