Automatically split and join a video using ffmpeg

November 28, 2021  [ffmpeg]  [bash] 

This summer I had a task of processing a large collection of videos that I recorded myself while teaching a course at NTNU. Most parts of the videos contained the content I wanted to preserve, but I wanted to cut away some small portions and, additionally, split the big video files into a series of smaller ones. Surely, I wanted to do this programmatically using ffmpeg. In this blog post I am going to summarize my experience with this task.

My overall workflow was quite simple: I was rewatching the videos in VLC and recorded timestamps I wanted to cut away in a text file. Once I was done with a portion of the original file that would further constitute a smaller clip with removed “bad” parts, I ran a bash script utilizimg ffmpeg to cut the big video in a set of smaller ones and then assemble them back in one single file.

My approach to recording timestamps was to create a line for each “good” small portion of the video I wanted to produce, delimited with a starting and ending timestamp im the format of HH:MM::SS:

01:26:10 01:44:04
01:44:28 01:44:55
01:45:23 01:56:39

Once such file was ready, I applied the following script:



mkdir "$tmp_dir"

echo "Video cutting and assembling: $video_file"
echo "Timestamps file: $timestamps_file"

echo "" > "$filelist"

while read -r line
    # Make array from the line (space-separated timestamps)

    # Cut video section with slower seek & no copying of codecs
    cmd="ffmpeg -y -i $video_file -ss $clip_start -to $clip_end $tmp_dir/$counter.mp4"

    echo "file '$counter.mp4'" >> "$filelist"


done < "$timestamps_file"

# Execute each command for cutting video sections
for cmd in "${commands[@]}"
    echo "$cmd"
    eval "$cmd"

# Merge clips into one file if more than one clip is available
if [ ${#commands[@]} -gt 1 ]; then
    ffmpeg -y -f concat -safe 0 -i "$filelist" -c copy "$tmp_dir/merged.mp4"

Let’s look at what the script does. First, it read two command line arguments, with the first ($1) containing the path to the video file and with the second ($2) storing the path to the file with the timestamps. Lazily enough, I created a temporary directory on Desktop where I was storing the intermediate files as well as a single resulting video file.

The timestamps file is parsed, and for each pair of the timestamps, an ffmpeg command is constructed to cut a small portion of the video:

ffmpeg -y -i $video_file -ss $clip_start -to $clip_end $tmp_dir/$counter.mp4

Each run of this command creates a file in the temporary directory starting with 0.mp4 and onward. Options of ffmpeg for this command are listed below:

Additionally, we are automatically populating a text file containing paths to the small video files we will later concatenate. For the example with three such file, filelist.txt will look as follows:

file 0.mp4
file 1.mp4
file 2.mp4

In the number of small files is greater than 1, they are concatenated using the following command:

ffmpeg -y -f concat -safe 0 -i "$filelist" -c copy "$tmp_dir/merged.mp4"

where $filelist stores the path to filelist.txt and the used options are the following:

As such, once I have the timestamps file (say, in the current directory), I invoke the script as follows ( stands for “video cut and assemble”): /path/to/bigvideo.mp4 timestamps.txt

After the processing is done, the temporary folder will contain the resulting file merged.mp4, along with the intermediate files that can be safely deleted.

Other nice tutorials about use cases of ffmpeg:

comments powered by Disqus