ffmpeg split screen two video sources

Question

I am not able to figure out how to get ffmpeg to make a video that combines (muxes) half-height-cropped from two video sources. The latest generation of ffmpeg (3.x) has a "vstack" filter.

My screen resolution is set to 1280 by 720 pixels.

I have a webcam that is set to provide video at that same resolution.

I want the top half of the webcam's input/feed/stream to be layered on top of the bottom half of the screen-capture-recorder (recording my desktop).

Both source video streams, plus the end output result will have a width of 1280 pixels. The combination of the half-height 2 video stream inputs will end up filling up 720 lines of vertical resolution.

This should be simple.

I'll paste a few of the commands that I've tried and that have failed to work.

ffmpeg.exe -f dshow -i video="screen-capture-recorder" -f dshow -i video="Microsoft LifeCam HD-6000" -vf "crop=1280:360:1280:360" -filter_complex vstack -vcodec libx264 -pix_fmt yuv420p -preset ultrafast 6000screen_take1.mkv

and

I even tried adding : "[0:v][1:v]vstack[v]" -map "[v]" to that , above (previous) command.

See my Google Doc for complete output (what was in my cmd.exe window) :

https://docs.google.com/document/d/1SqtjEY_gM3rdeKPyRzUq_jJhPQUPCK5L4E5HPttwS3s/edit?usp=sharing

Answer 1

If you're using filter_complex, then all filtering has to happen within that complex. So,

ffmpeg.exe -f dshow -i video="screen-capture-recorder" \
           -f dshow -i video="Microsoft LifeCam HD-6000" \
-filter_complex "[0:v]crop=1280:360:0:0[v0];[1:v]crop=1280:360:0:0[v1];[v0][v1]vstack[v]" \
-map [v] -vcodec libx264 -pix_fmt yuv420p -preset ultrafast 6000screen_take1.mkv

This stacks the top half of the webcam video below the top half of the desktop capture.

---

Having taken a look at the console output, you should actually try this:

ffmpeg.exe -f dshow -rtbufsize 9000000 -i video="screen-capture-recorder" \
           -f dshow -i video="Microsoft LifeCam HD-6000" \
-filter_complex "[0:v]crop=1280:360:0:0[v0];[1:v]scale=1280:-1,crop=1280:360:0:0[v1]; \
                 [v0][v1]vstack[v]" \
-map [v] -vcodec libx264 -pix_fmt yuv420p -preset ultrafast 6000screen_take1.mkv

Answer 2

I made another video and posted it to YouTube.

I used the following command :

  ffmpeg.exe -f dshow -i video="screen-capture-recorder" -f dshow -i video="Microsoft LifeCam HD-6000":audio="virtual-audio-capturer" -filter_complex amix=inputs=1 -filter_complex "[0:v]crop=1280:360:0:0[v0];[1:v]scale=1280:-1,crop=1280:360:0:0[v1]; [v0][v1]vstack[v]" -map [v] -vcodec libx264 -pix_fmt yuv420p -preset ultrafast -acodec pcm_s16le (output video file('s name))

Notice how what is depicted on the top half (virtual on-screen piano keyboard) (both of them) lags behind. Yeah there is latency between the video signal of the top half and the rest of the content.

I suppose there is a more complex solution to this.

I tried tying (chaining) the :"audio=" to both of the video input source/streams

that is , I tried both logical possibilities (variations) :

video="Microsoft LifeCam HD-6000":

and

video="screen-capture-recorder":

Same issue. In fact, the latency doesn't exchange between the different video streams.