How to Trim and Reencode Video Files
Here are FFmpeg commands which we have found useful for making clips from video files. We also show you how to reencode the video. You may need to do this in order to make the smaller or to get the video into a format which your player can play.
Trim a Video File
These commands both extract an eight-second piece out of movie.mp4 starting three seconds in and save it as clip.mp4:
$ ffmpeg -i movie.mp4 -ss 00:00:03 -t 00:00:08 -c copy clip.mp4
$ ffmpeg -i movie.mp4 -ss 00:00:03 -to 00:00:11 -c copy clip.mp4
Here is what each parameter means:
-i the name of the input file
-ss the start time of the clip
-t the duration of the clip
-to the end time of the copy
-c copy copy the audio and video information unchanged. This preserves the quality and leaves the compression method the same.
Because the command above does not reencode the video, it has to include additional material at both the start and the end until it reaches the next keyframe. It then inserts instructions telling the player not to show this material. This hidden material may be several seconds long. If you need to make sure nobody can possibly see this extra material or you will be using a player which does not handle this situation properly, you will need to reencode the video.
Reencode a Video File (H.264 codec)
As of January 2020, one of the most popular video formats is MPEG-4 with the H.264 video codec and the AAC audio codec. This will play on most computers, smartphones, and inexpensive standalone devices which play video from a thumbdrive.
The general form of the command to convert video to this format is:
$ ffmpeg -i movie_in.mp4 _video options_ _audio_options_ movie_out.mp4
Here are some video options which will produce a high-quality result:
-c:v libx264 -profile:v high -level 4.1 -preset slow -pix_fmt yuv420p -crf 25
Here is what each parameter means:
-c:v use the H.264 video codec.
-profile:v high -level 4.1 tells the H.264 encoder it is ok to use advanced compression methods, our player can deal with it.
-preset slow tells the H.264 encoder to take its time and do a good job. It may take a few hours to encode an hour-long video. If you cannot wait, change this to -preset fast.
-pix_fmt yuv420p tells FFMpeg how to encode the color information. If you leave this out, and the input file uses some other encoding, some hardware players may be unable to play the file produced.
-crf 25 sets the level of compression. Higher numbers mean more compression and more loss of detail. Noticable smearing sets in around 35.
H.264 video is usually paired with AAC audio. These audio options are good, but they require an FFmpeg compiled from source:
-c:a libfdk_acc -profile:a aac_he -ar 48000
These options are known to work with the FFmpeg supplied with Ubuntu 18.04:
-strict experimental -c:a aac -ar 48000
Reencoding a Video File (VP9 codec)
VP9 is a newer codec from Google. It will play in recent releases of VLC and in most web browsers, but not on most standalone hardware devices. VP9 video is usually combined with Opus audio in a Webm container.
The general form of the command to produce a Webm file is:
$ ffmpeg -i movie_in.mp4 _video options_ _audio_options_ movie_out.webm
These video options will produce a high-quality result:
-c:v libvpx-vp9 -speed 1 -pix_fmt yuv420p -crf 35
And here are reasonable audio options:
-c:a libopus -ar 48000
Reencoding a Video File (MPEG-4 Part 2)
On a typical desktop computer circa 2018 encoding a video using the H.264 codec as described above will take about two minutes for every minute of video. The VP9 codec can take even longer. During this time the computer is working hard searching for a way to make the file as small as possible while preserving quality. This is important when you will be distributing your videos on disks or on the Internet since more of them fit and they can be downloaded more quickly.
But sometimes you need a video encoded right away such as to show at a class which will start in a few minutes. At these times you can use an obsolete codec from when computers were less powerful.
A good choice for this is the MPEG-5 Part 2 codec. These video options are a good starting point:
-c:v mpeg4 -qscale:v 7
On a modern computer this will encode an hour-long video in about 15 minutes. It will be nearly twice as large as the same video in H.264, but that is the price you pay for speed.
qscale:v sets the level of compression. A lower number gives higher quality video, a higher number produces smaller files.
While in the past MPEG-4 Part 2 video was often paired with MP3 audio, for this application it is reasonable to pair it with AAC audio and put them in an MP4 container like with did for H.264. See above for the AAC options.
Scaling the Video
To reduce the resolution of the image, use the scale video filter. You may want to do this if you now that the video will be played in a small window or on a device with limitted resolution. Cutting the resolution will allow the file to be much smaller and it may play more smoothly on old devices.
For example, if you have 1280x720 video and want to reduce it to 640x360, add this filter to the ffmpeg command:
-vf scale=640:-1
Here we specify the target width and allow FFmpeg to computing the cooresponding new height. You can add this option to any of the commands above.
Source File from HTTP
If the source file is on a website, you may not need to download the entire file to take a clip. Simply use the URL as the source file name. FFmpeg will download only the parts of the file which are needed. (Presumably the server must support HTTP range requests.)
