Why not wrap avconv and ffmpeg using the Python sh module[1]?
I was kind of curious about the problem myself so I did an
implementation of the thumbnail extraction using ffprobe and ffmpeg
(attached).[2]
[1] python-pbs pre-jessie, python-sh in jessie and sid
https://github.com/amoffat/sh
[2] based on this answer from SU: http://superuser.com/a/821680
On Sat, Mar 21, 2015 at 3:20 AM, Etienne Millon <m...@emillon.org> wrote:
> * Brian May <br...@microcomaustralia.com.au> [150321 11:07]:
>> I need to be able to generate a thumbnail of designated size and
>> also tell what the width and height in pixels of the video; Being
>> able to scale the video would be even better, but optional, I
>> currently use avconv to do this.
>
> You should be able to get metadata using python-enzyme ; not sure
> about what's best to actually extract the thumbnail.
>
> --
> Etienne Millon
>
>
> --
> To UNSUBSCRIBE, email to debian-python-requ...@lists.debian.org
> with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
> Archive: https://lists.debian.org/20150321102055.GA3789@klow
>
from sh import ffmpeg,ffprobe
import sys
import json
movie = sys.argv[1]
frames = int(sys.argv[2])
videometadata = ffprobe("-v", "quiet",
"-print_format", "json",
"-show_format",
"-show_streams",
movie)
videometadata = json.loads(videometadata.stdout)
duration = float(videometadata['format']['duration'])
print "Generating %d thumbnails for %s" % (frames,movie)
for x in range(1,frames+1):
timepoint = int( (x - 0.5) * duration / frames)
print " Extracting thumbnail %d/%d @ %d" % (x,frames,timepoint)
ffmpeg('-y', '-v', 'quiet',
'-ss', timepoint,
'-i', '%s' % movie,
'-vframes', '1',
'image%d.jpg' % x)
