Python script to enforce a maximum duration on a set of audio files. Long files are trimmed, without altering the sample data. The script can either convert all to FLAC, or preserve exact container type for each file.

trim_audio_dataset.py

#!/bin/env python

# Script to enforce a maximum duration on a set of audio files.
# Long files are trimmed, without altering their file format.
# Uses ffmpeg. (So, that needs to be installed on your system).
# Tested in Ubuntu Linux v22.04.

# Written by Dan Stowell 2024.
# CC0: This work has been marked as dedicated to the public domain.

import os, glob, subprocess

###################################################
# User config variables:
indir = "wavs_original"
outdir = "wavs_trimmed"
maxdur = 120      # measured in seconds
verbose = True

convert_format = True     # whether to convert all to FLAC (should achieve smaller filesize without data loss)

###################################################

totalin = 0
totalout = 0

for inpath in sorted(glob.glob(os.path.join(indir, "*.*"))):
    filename = os.path.basename(inpath)
    outpath = os.path.join(outdir, filename)
    if convert_format:
        acodec = 'flac'
        outpath = os.path.splitext(outpath)[0] + ".flac"
    else:
        acodec = 'copy'
    totalin += 1
    if verbose:
        print("======================")
        print(f"File {filename}:")

    # First calculate the duration
    cmd = ["ffprobe", "-v", "error", "-show_entries", "format=duration", "-of", "default=noprint_wrappers=1:nokey=1", inpath]
    try:
        result = subprocess.run(cmd, capture_output=True, check=True, text=True)
        #print(result)
        filedur = float(result.stdout)
    except Exception:
        print(f"WARNING: Unable to read duration of input file: {filename}")
        if verbose:
            print(result.stdout)
            print(result.stderr)
        continue

    # This is the main command. We will add more arguments.
    cmd = ["ffmpeg", "-loglevel", "quiet", "-y",
        "-i", inpath, "-acodec", acodec]

    if filedur <= maxdur:
        # Note that despite "no trim needed", we still use ffmpeg to copy it,
        # to ensure the file's container is readable,
        # i.e. all output files will surely be readable at least by ffmpeg.
        if verbose:
            print(f"File will be kept full length: {filename}, full dur {filedur}")
        pass

    else:
        if filedur <= maxdur * 3:
            # choosing the middle segment
            starttime = (filedur - maxdur) / 2
        else:
            # choosing the segment [maxdur, maxdur * 2]
            starttime = maxdur
        if verbose:
            print(f"File will be trimmed: {filename}, full dur {filedur}, start will be {starttime}")
        cmd += ["-ss", str(starttime), "-t", str(maxdur)]

    cmd += [outpath]  # the command is finally completed, with the output path
    if verbose:
        print(cmd)
        print(" ".join(cmd))
    try:
        result = subprocess.run(cmd, capture_output=True, check=True, text=True)
    except Exception:
        print(f"WARNING: Error creating edited file for: '{filename}'")
        if verbose:
            print(result.stdout)
            print(result.stderr)
        continue
    totalout += 1

print(f"Finished. Converted {totalout} out of {totalin} files.")