ffmpeg concat and scale simultaneously? - linux

Question

I have two ffmpeg commands:
ffmpeg -i d:\1.mp4 -i d:\1.mp4 -filter_complex "[0:0] [0:1] [1:0] [1:1] concat=n=2:v=1:a=1 [v] [a]" -map "[v]" -map "[a]" d:\3.mp4
and
ffmpeg -i d:\1.mp4 -vf scale=320:240 d:\3.mp4
How to use them both simultaneously?

Answer 1

For posterity:
The accepted answer does not work if the input sources are of different sizes (which is the primary reason why you need to scale before combining).
What you need to do is to first scale and then pipe that video output into the concat filter like so:
ffmpeg -i input1.mp4 -i input2.mp4 -filter_complex \
"[0:v]scale=1024:576:force_original_aspect_ratio=1[v0]; \
[1:v]scale=1024:576:force_original_aspect_ratio=1[v1]; \
[v0][0:a][v1][1:a]concat=n=2:v=1:a=1[v][a]" -map [v] -map [a] output.mp4
Had this problem today and was pulling my hair for good three hours trying to figure this out and unfortunately the accepted answer did not work as noted in the comments.

Answer 2

ffmpeg -i d:\1.mp4 -i d:\2.mp4 -filter_complex "concat=n=2:v=1:a=1 [v] [a]; \
[v]scale=320:200[v2]" -map "[v2]" -map "[a]" d:\3.mp4
Firstly we concatenate everything and pipe result to [v] [a] (see filtergraph syntax docs - its output from concat filter). Next we take [v], scale it and output to [v2], lastly we take [v2] and [a] and mux it to d:\3.mp4 file.

Answer 3

Construct custom filtergraph, move resize procedure nearer to the video source, for example let's deal with a more complex graph in order to grasp the spirit of its construction language:
ffmpeg.exe -i Movie_oriented_minus_90.mov -i Movie_pause.mp4 -i Sound_pause.aac -filter_complex "[0:v:0]scale=1920:1080 [c1],[c1]vflip[c2],[c2]hflip[clip], [clip] [0:a:0] [1:v:0] [2:a:0] concat=n=2:v=1:a=1 [v] [a]" -map "[v]" -map "[a]" -c:v libx264 -q:v 0 -acodec mp3 -s 1920x1080 Movie_oriented_plus_90_with_pause.mp4
Stage 1: Movie_oriented_minus_90 is source 0, its video stream
[0:v:0] fed into scale filter and produced as [c1], then [c1] flipped
vertically into [c2] and then [c2] flipped horizontally into [clip]
thus rotated for 180 degree
Stage 2: 1st video stream concatenated with 2nd source stream, i.e.
1st video stream: [clip] (processed stream from source 0) and sound
from original video [0:a:0] 2nd video stream: constructed from video
from source 1 [1:v:0] and audio [2:a:0] from source 2 (30 sec of
silence made with -filter_complex "aevalsrc=0:d=30" during separated
run of ffmpeg)
Stage 3: the resulting video sequence [v] and [a] then compressed
with x264 codec into the target mp4 file
So, the main problem with your question was that you tried to concatenate streams with dfferent sizes and only then applied resizing operation for the already aggregated stream which is of course can't consist of media samples with different sizes.