The core command for those running ffmpeg on a Unix-compatible system (e.g. MacOS, BSD and GNU-Linux) is really quite simple. It's to redirect or to "pipe" one of the outputs of ffmpeg to ffplay. The main problem here is that ffmpeg cannot autodetect the media format (or container) if the output doesn't have a recognizable file extension such as .avi or .mkv.
Therefore you should specify the format with the option -f. You can list the available choices for option -f with the ffmpeg -formats
command.
In the following GNU/Linux command example, we record from an input source named /dev/video0
(possibly a webcam). The input source can also be a regular file.
ffmpeg -i /dev/video0 -f matroska - filename.mkv | ffplay -i -
A less ambiguous way of writing this for non-Unix users would be to use the special output specifier pipe
.
ffmpeg -i /dev/video0 -f matroska pipe:1 filename.mkv | ffplay -i pipe:0
The above commands should be enough to produce a preview. But to make sure that you get the video and audio quality you want, you also need to specify, among other things, the audio and video codecs.
ffmpeg -i /dev/video -c:v copy -c:a copy -f matroska - filename.mkv | ffplay -i -
If you choose a slow codec like Google's AV1, you'd still get a preview, but one that stutters.