Skip to content

Audio Encoding

Available Encoders

There are plenty with varying degrees of usefulness.



Available Extractors/Trimmers


Eac3to is at this point fully obsolete because of our ffmpeg implementation.
There's only really a use for it if you need to add silence to something proprietary like TrueHD Atmos or DTS:X without reencoding.


SoX only works with lossless files and always uses wav internally.
It also is more accurate than FFMPEG and as such should be your default for lossless.
FFMPEG can somewhat decently trim and concat lossy audio and we implemented something similar to the eac3to delay thing to have a more accurate end result.

Basic Usage

All-in-one Function

There's this neat little do_audio function that should satisfy all your needs.
If you don't pass any extractors, trimmers or encoders it will choose:

  • FFMPEG as the extractor

  • SoX or FFMPEG for the trimmer depending on if the input is lossy (the latter if it is)

  • Opus as the encoder if lossless and otherwise it will choose to not encode

If you already have wav files to work with you can also pass None to the extractor param to skip that step.

from vsmuxtools import src_file, Wavpack, do_audio # Another alias for this would be encode_audio

file = R"00008.m2ts"
audio = do_audio(file, 1, encoder=Wavpack())

# You can also pass a src_file with the vs(!)-muxtools extension.
# This will automatically apply the src_file trim if you don't pass any yourself.
ITBD = src_file(file, trim=(24, -24))
audio = do_audio(ITBD, 0)

# If you want to choose everything
from vsmuxtools import FFMpeg, Sox, TrueAudio

audio = do_audio(JPBD, track=0, extractor=FFMpeg.Extractor(preserve_delay=False), trimmer=Sox(), encoder=TrueAudio())

Step by step

Of course you're not forced to use do_audio.

from vsmuxtools import FFMpeg, Sox, FLAC, DitherType, src_file, Resample

ITBD = src_file(R"00008.m2ts", trim=(24, -24))

audio = FFMpeg.Extractor(track=1).extract_audio(ITBD.file)
audio = Sox(trim=[(24, 500), (800, 1245)], fps=Fraction(24000, 1001)).trim_audio(audio)
audio = FLAC(preprocess=Resample(dither=DitherType.LOW_SHIBATA)).encode_audio(audio)

All of those audio variables are a AudioFile object.
To get the filepath you simply do audio.file.

Same deal as with the encodes these also have a to_track() function to create a track with metadata. (Defaults to lang='jp', default=True, forced=False)


(Almost) every encoder also supports various preprocessing options.
Currently implemented are:

  • Resample
    To change sample rate or bitdepth

  • Downmix / Pan
    The ffmpeg pan filter with a few presets for better/more dynamic downmixing

  • Loudnorm
    A ffmpeg based normalization implementation according to EBU-R128 standards.
    This also uses the two-pass methodology as described here and used in the ffmpeg-normalize utility.

Example usage

# Lossless encoders use the Resample preprocessor by default (which defaults to 16bit and 48kHz)
audio = do_audio(file, 1, encoder=FLAC())

# Can be entirely disabled like this
audio = do_audio(file, 1, encoder=FLAC(preprocess=None))

# You can also chain them like this
audio = do_audio(file, 1, encoder=FLAC(preprocess=[Pan(Pan.RFC_7845), Resample(DitherType.LIPSHITZ)]))

# Or define your own with either ffmpeg args or an ffmpeg audio-filter
audio = do_audio(file, 1, encoder=FLAC(preprocess=[
    # In this case, the filter creates a massive echo effect (the args are obviously nonsense)
    CustomPreprocessor(filt="aecho=0.8:0.9:1000:0.3", args=["--example", "value"])