Discover the Six FFmpeg Commands You Can’t Live Without
FFMPEG is a free command-line utility that serves as the engine for most of the largest cloud encoding farms in the world, public and private. But it also performs many simple and essential tasks that crop up in a studio or encoding facility, particularly if you do performance or benchmark testing. In this article, I’ll show you how to perform six of these tasks.
To get started, you can download FFmpeg for free from FFmpeg.org/download.html, with packages available for Windows, Mac, and multiple Linux flavors including Ubuntu, Debian, Fedora, and Red Hat Enterprise Linux. Though batch command structure and operation is different for each platform, the FFmpeg command syntax is very similar for all.
Here’s the basic FFmpeg command line structure:
ffmpeg -i input.mp4 output.mp4
Here’s an explanation for the commands:
This calls the program; add drive and folder information for ffmpeg.exe if it’s not in your path.
This identifies the input file.
This is the input file. FFmpeg can input most container formats natively, including MP4, .ts, MOV, AVI, Y4M, MKV, and many others.
This is the output file. No switch necessary; just the file name.
Note that unless you identify an audio or video codec via switches, FFmpeg will use the default codecs for each container format as well as reasonable encoding parameters. For the MP4 extension, if you input a 1080p file, FFmpeg will encode using the H.264 video codec at about 9 to 10 Mbps, the AAC audio codec at around 130 Kbps, a keyframe interval of 250 frames, the High profile, and the medium x264 preset. All of these are customizable, of course, but if all you’re trying to do is to create a video file you can play from a hard drive, the command above will suffice.
Note that you can download all commands and the input test file used in this tutorial. (Note that the download will begin automatically when you click this link.) The video is excerpted from a test video provided by Harmonic, Inc. (harmonic inc.com).
Changing Container Formats Without Re-Encoding
Say you’re working with an application that only accepts MOV files and your source file is in an MP4 container (or vice versa). You could input the file into Adobe Premiere Pro and output as needed, but Premiere Pro can’t do that without re-encoding. This is easily done in FFmpeg with this command string:
ffmpeg -i input.mp4 -c:v copy -c:a copy output.mov
The new commands here are as follows:
Video codec, copy. This tells FFmpeg to copy the compressed video stream into the new file without re-encoding.
Audio codec, copy. This tells FFmpeg to copy the compressed audio stream into the new file without re-encoding.
A similar scenario might involve files in the MPEG-2 container format with a .ts extension. Here you may have to add one switch to handle the MPEG-2/4 ADTS header from the transport stream file.
ffmpeg -i input.ts -bsf:a aac_adtstoasc -c:v copy -c:a copy outputfromts.mp4
The added switch is as follows:
manages the ADTS conversion.
This particular conversion saved about 2 days of encoding when I had to convert about 100 .ts files into MP4 files for testing. The
aac_ adtstoasc switch may not be necessary with more recent versions of FFmpeg, which may insert the switch automatically. You’ll know you need it if the conversion crashes or the audio doesn’t play normally in the converted file. Or, you can just leave it in.
Extracting a Section of a File Without Re-Encoding
Say you have a file and want to eliminate the first 30 seconds, or simply grab a 2-minute excerpt from the middle. Again, you could bring the file into Adobe Premiere Pro and edit what you want, but that would involve another rendering and encoding cycle. You can do this quickly and simply in FFmpeg using the command shown below.
The Harmonic test file is about 12 seconds long. This command seeks to 5 seconds into the file and excerpts the next 3 seconds without re-encoding.
ffmpeg -ss 00:00:05 -i input.mp4 -t 00:00:03 -c:v copy -c:a copy excerpt.mp4
Seeks to 5 seconds into the file.
Extracts this duration of the file. If you leave this switch off, FFmpeg will include from the seek point to the end of the file in the extracted file.
Note that FFmpeg seems to work around keyframes in the file so the results are typically not frame-accurate. This doesn’t matter in most instances, but if you absolutely need a specific duration or specific frames included or excluded from your excerpted file, this technique probably won’t work. In these cases, you’re better off using your video editor.
If you want to extract only the video in the file and not the audio, add the
-an switch. To extract audio only, add the
-vn switch. Note that the MP4 container can hold audio-only files, so you don’t need to change container formats when producing audio-only files from MP4 sources.
Converting to a Raw Format
Some file analysis tools only input raw YUV or Y4M files, so you’ll have to convert your encoded files to YUV for input. In addition, when working with newer codecs like AV1, most tools can’t input the encoded file natively; again, you’ll have to convert to YUV or Y4M.
What’s the difference between YUV and Y4M? Both are uncompressed formats, but the YUV file header doesn’t contain resolution, frame rate, or other file-related information. To input a YUV file into a program, you’ll typically need to specify one or more of these parameters. In contrast, Y4M files contain this data, simplifying input.
For example, to input a YUV file like ConstructionField.yuv into the Moscow State University Video Quality Measurement Tool, you have to specify resolution and color information as shown at the top of Figure 1. With a Y4M file, this data is in the file header, so you can skip this step, both in the user interface and in the batch command string. Basically, if the tool you’re working with can accept Y4M files, use that format; if not, use YUV.
Figure 1. Inputting a YUV file into the Moscow State University Video Quality Measurement Tool forces you to specify resolution and other file- related data. Y4M files contain this in the header and input normally
Converting to either format is simple using this command:
ffmpeg -i input.mp4 -pix_fmt yuv420p -vsync 1 output.y4m
To produce a YUV file, simply use that extension. Here’s an explanation for the new commands:
This sets the file format, and YUV420p is the most commonly used format for 8-bit files. If you’re working with 10-bit formats, check out this conversation on Stack Exchange.
This tells FFmpeg to preserve the audio/video sync between the input and output files. I’m not 100% sure it’s necessary in every instance, but long ago it cured a sync issue for me, and I’ve left it in ever since.
Scaling Compressed Files When Converting to YUV/Y4M
This is a very common use case when computing objective quality metrics on files compressed at lower resolutions than the original. For example, imagine you wanted to compute the VMAF rating on a 720p file encoded from a 1080p file. With most tools, you’ll have to scale the encoded file back to 1080p so you’ll have a pixel-by-pixel match with the source file.
To do this, use the same basic command as shown above, adding two new switches.
ffmpeg -i input.mp4 -s 1920x1080 -sws_flags lanczos -pix_fmt yuv420p -vsync 1 output.y4m
Here’s an explanation for the new commands:
This tells FFmpeg to scale back to 1080p.
This tells FFmpeg to use the Lanczos method of scaling
I use this approach because it is the primary scaling technique used by NVIDIA when scaling video from lower to higher resolutions, and because when performing quality measurements, I’m trying to predict video quality as viewed by someone watching on their computer or on a smart TV. Though there are higher-quality scaling methods, they’re probably too slow for use on most display devices, so I stick to Lanczos.
That’s it. If you’ve never worked in the command line, you’ll definitely find yourself at a disadvantage, but an hour or two invested in learning basic command line will get you up-to-speed. If you frequently experiment with encoding and quality metrics, you’ll find the command line and FFmpeg absolutely essential, and that the combination vastly simplifies and accelerates your testing and analysis workflows.
[This article appears in the June 2019 issue of Streaming Media Magazine as "Six FFmpeg Commands You Can’t Live Without."]
With just a few beginner-level scripts, you can encode and package multiple filds to HLS and DASH output using open source tools.
AV1 delivers equivalent quality to HEVC, but with a lower data rate. For now, though, it's slow. A five-second clip took 23 hours and 46 minutes to encode.
Building on previous DIY articles, this installment will walk you through capturing audio, encoding and packaging with FFmpeg, then playing out through the Icecast server you set up in the last article
FFmpeg might be a free download, but there's still a cost that goes with it. Robert Reinhardt explains why this encoder isn't right for every company.