Timelapse without using storage for jpegs: pipe the raspistill output directly to mencoder

Question

I want to make a timelapse with my Pi, but rather than storing thousands of images on my small SD card and then spend hours mencoder'ing the result, I'd like to pipe the raspistill -tl output (if possible) into mencoder, and get the resulting compressed video when I stop the process.

Since I plan to take 1 picture every 5s, I figure there's plenty of time to append that image to the mencoder video between shots.

Is is possible?

I tried a simple script but have an error:

$ raspistill -a 12 -tl 1000 -t 10800000 -o - | mencoder - -nosound -ovc lavc -lavcopts vcodec=mpeg4:aspect=16/9:vbitrate=8000000 -vf scale=1920:1080 -o timelapse.avi -mf type=jpeg:fps=30
MEncoder SVN-r37379 (C) 2000-2015 MPlayer Team
Reading from stdin...
success: format: 0  data: 0x0 - 0x0
libavformat version 56.25.101 (internal)
Cannot seek backward in linear streams!
Seek failed
Cannot seek backward in linear streams!
Seek failed
Cannot seek backward in linear streams!
Seek failed

I suppose a list of JPG is not exactly what mencoder gets from the input. Any guru here to help me?

edit

With a separate command invocation:

raspistill -a 12 -tl 1000 -t 10000 -o - > test.raw # takes 10 images across 10 s
mencoder test.raw -nosound -ovc lavc -lavcopts vcodec=mpeg4:aspect=16/9:vbitrate=8000000 -vf scale=1920:1080 -o timelapse.avi -mf type=jpeg:fps=30

This produces a valid video, so there's something in the pipe that doesn't work correctly.

Answer 1

I was trying to do the exact same thing. After fiddling around with terminal piping and getting rather mixed results (incomplete files, dropped frames, large filesize, etc.), I found that the Python picamera module (preloaded on Raspbian) can give excellent H.264 timelapse videos easily.

For the Camera module v1, PiCamera allows setting the video framerate as low as one frame per 6 seconds. For the Camera module v2, the framerate can go down to one frame per 10 seconds. These should cover most use cases.

The bonus is that all processing is done on the GPU, so this requires almost no CPU time.

Example:

from picamera import PiCamera
from fractions import Fraction
camera = PiCamera()
try:
        camera.resolution = (1920, 1080)  # 1080p is the max resolution of the H.264 encoder
        camera.framerate = Fraction(1, 3) # One frame every 3 seconds
        camera.start_recording('output.mp4', format='h264', quality=23)
        camera.wait_recording(300)        # Time of recording, in seconds
        camera.stop_recording()
finally:
        camera.close()