The ffmpeg command line tool is a hyper-fast video and audio converter that can also grab from a live audio/video source. It can convert between arbitrary sample audio/video rates and resize/trim videos on the fly with a high-quality polyphase filter.

Suppose you want to fast-forward this masterpiece by 2x:-

Fast Forward (Video Only)

When you execute the above with the correct inputs, the output file will contain the fast-forwarded video file without the audio stream.

However, if you still want to fast-forward both audio and video simultaneously, you must use a complex filter. I will teach you how we can do that in the end (TAKE ME THERE!). But first, let's understand what the above command does.

Behind the Scenes

-i hehe-peter.mp4

The string hehe-peter.mp4 followed by the -i flag describes location of the source input file. Conversely, the last argument for the command is the name/target file name for the output.

-filter:v "setpts=0.5*PTS"

The "setpts" filter changes the "PTS" (presentation timestamps) of video frames. This is the filter we use to change the frame rate and do slow-motion or fast-motion effects. The filter accepts an expression that specifies the new value of the PTS. For example, setpts=0.5*PTS will double the frame rate, effectively making the video play at 2x speed.

-preset fast

This argument specifies the speed-to-compression ratio. ffmpeg has a collection of options providing certain encoding speeds such as ultrafast, superfast, veryfast, faster, fast, medium (default), slow, slower, and veryslow.

A slower preset will provide better compression (compression is quality per file size). For example, if you target a specific file size or constant bit rate, you will achieve better quality with a slower preset. Similarly, choosing a slower preset will save bitrate for constant quality encoding. Use the slowest preset that you have patience for, haha.

-crf 24

Constant Rate Factor configuration or crf is a procedure that allows the encoder to attempt to achieve a certain output quality for the whole file while controlling the output file size. A lower CRF value results in higher quality but larger file size.

The values for this argument will depend on which encoder you're using. For example, I'm using the x264 encoder like in the above command. Hence, the possible value is an integer from 0-51.

-c:v libx264

-c:v libx264 specifies the codec to be used for the video stream of the output file. The "v" in "c:v" stands for video. The libx264 is the codec name, it's a codec library for H.264 video compression. It tells ffmpeg to use the libx264 codec to encode the video stream in the output file. Also, -vcodec and -codec:v is an alias for -c:v.

You can see the supported codecs list by running ffmpeg -codecs. Or, if you want a more fine-grained list, you can see encoders or decoders using ffmpeg -encoders or ffmpeg -decoders. And see more detailed information using ffmpeg -h encoder=libx264 or ffmpeg -h decoder=aac. Similarly, you can do the same for listing the formats supported by ffmpeg using ffmpeg -formats.

-an

In ffmpeg, the -an option tells ffmpeg to disable the audio stream in the output file. It stands for "audio disable". Alternatively, we also can:-

  • Use the -vcodec copy option: This option tells ffmpeg to only include the video stream in the output file. For example, ffmpeg -i input.mp4 -vcodec copy output.mp4

  • Use the -map option: This option allows you to specify which streams from the input file should be included in the output file. For example, ffmpeg -i input.mp4 -map 0:v -c:v libx264 output.mp4. This will only include the first video stream from the input file in the output file.

  • Use the -codec:a none instead of the -an option *.

When you include one of these flags in a command, ffmpeg will not process any audio stream from the input file, and it will not include any audio stream in the output file.

Well, what if you want to fast-forward both audio and video streams? would appending -acodec copy work? Let's see what's the outcome.

Did you see? The video is fast-forwarded but the audio isn't. And we are still left with the original duration. Why? you may ask.

Fast Forward (Video & Audio)

Executing the above command with the correct input and a targeted output file will yield the fast-forwarded video and audio.

And would you look at that! I've just ruined that masterpiece, lol. But seriously though, I think ffmpeg is pretty cool because it has all these fancy features :)

Behind the Scenes (Cont.)

-filter_complex "..."

In ffmpeg, the -filter-complex option specifies a complex filtergraph. A filtergraph is a directed acyclic graph of connected filters. Each filter takes one or more inputs, processes them, and produces one or more outputs.

The above command has two filter graph nodes separated by a semicolon. [0:v]setpts=0.5*PTS[v] is what allows us to fast-forward the video by 2x speed. In this argument, [0:v] tells ffmpeg which video stream of the input file to use. "0" is the index of the input file, and "v" specifies that we want to use the video stream.

Note that we use the atempo filter to change tempo of the audio. However, the value of it are limited to using values between 0.5 and 2.0 (so it can slow it down to no less than half the original speed, and speed up to no more than double the input). If you need to, you can get around this limitation by stringing multiple atempo filters together (i.e., -filter:a "atempo=2.0,atempo=2.0").

Essentially, where 0:v specifies the first video stream of the input file as the input link for the filter setpts. And the [v] appended at the end is an output link specification in filtergraphs. It tells ffmpeg to use the output of a filter as the input for the next filter in the chain or as the final output if it's the last filter in the chain.

-map "..."

In ffmpeg, the -map "[v]" option specifies which streams from the input file should be included in the output file. It tells ffmpeg to map the output link set inside the brackets (in this case, "[v]" maps the video and "[a]" maps the audio) to the output file.

We often use it with filtergraphs, where there can output of the filtergraph is used as the input for the mapping. We can use the -map option multiple times to select various streams from the input files and to specify the order in which they will appear in the output file. Also, note that if you want to use the entire input file, you can use -map 0 instead of -map "[v]".

If you need a more coarse-grained guide on how to slow-motion or fast-forward videos, you can refer to the official guide on ffmpeg wiki. But for me, ffmpeg documentation is the primary and most reliable source of information on the tool's features and capabilities.

Well, that's it, folks. Thanks a bunch for reading!

Published