Creating thumbnails with FFmpeg

Question

I am using FFmpeg to extract thumbnails from specific positions of video files.

I found on the web two approaches to do this:

1. With -ss (seek) parameter before -i (input) parameter:

   ffmpeg -y -ss $SEEK_POINT -i input.ogv -vcodec mjpeg -vframes 1 -an -s 120x90 -f rawvideo output.jpg

2. With -ss (seek) parameter after -i (input) parameter:

   ffmpeg -y -i input.ogv -vcodec mjpeg -ss $SEEK_POINT -vframes 1 -an -s 120x90 -f rawvideo output.jpg

The first method generates a bad thumbnail with gray spots, but works very fast. The error returned is [theora @ 0x8097240] vp3: first frame not a keyframe.

The second method always works but show an error which cause the extraction to take a lot of time. This amount of time is not fixed and it depends on the seek point as I noticed. Sometimes it takes a few seconds and other times several minutes to extract a thumbnail. I get the error Buffering several frames is not supported. Please consume all available frames before adding a new one. in the following output:

Input #0, ogg, from 'input.ogv':
  Duration: 00:21:52.76, start: 0.000000, bitrate: 844 kb/s
    Stream #0.0: Video: theora, yuv420p, 800x600 [PAR 4:3 DAR 16:9], 25 fps, 25 tbr, 25 tbn, 25 tbc
    Stream #0.1: Audio: vorbis, 44100 Hz, stereo, s16, 192 kb/s
    Metadata:
      ENCODER         : Lavf52.102.0
Incompatible pixel format 'yuv420p' for codec 'mjpeg', auto-selecting format 'yuvj420p'                                                                         
[buffer @ 0x9250840] w:800 h:600 pixfmt:yuv420p                                 
[scale @ 0x92508a0] w:800 h:600 fmt:yuv420p -> w:120 h:90 fmt:yuvj420p flags:0x4
Output #0, rawvideo, to 'output.jpg':
  Metadata:
    encoder         : Lavf53.2.0
    Stream #0.0: Video: mjpeg, yuvj420p, 120x90 [PAR 4:3 DAR 16:9], q=2-31, 200 kb/s, 90k tbn, 25 tbc
Stream mapping:
  Stream #0.0 -> #0.0
Press ctrl-c to stop encoding
[buffer @ 0x9250840] Buffering several frames is not supported. Please consume all available frames before adding a new one.                                    
frame=    0 fps=  0 q=0.0 size=       0kB time=10000000000.00 bitrate=   0.0kbit
Last message repeated 15448 times
frame=    1 fps=  0 q=3.4 Lsize=       3kB time=0.04 bitrate= 598.8kbits/s    
video:3kB audio:0kB global headers:0kB muxing overhead 0.000000%

How can I extract thumbnails without any problems using FFmpeg from a custom position of a video regardless of the input format?

Solution

Try something like this with the itsoffset option:

ffmpeg -itsoffset 4 -i "$INFILE" -vcodec png -vframes 1 -an -f rawvideo -s 120x90 -y "$OUTFILE"

From FFmpeg documentation on itsoffset:

Set the input time offset in seconds. [-]hh:mm:ss[.xxx] syntax is also supported. The offset is added to the timestamps of the input files. Specifying a positive offset means that the corresponding streams are delayed by offset seconds.

Comments

[Calin-Andrei Burloiu]

No, it does not work. Every time I run, it only extract the very first frame from the video, regardless of the parameter I set to -itoffset. Do you have any idea why?

[runfalk]

Late reply for anyone finding this later on. I believe the parameter -itsoffset should be negative. In this case -4 since you want to move the video stream earlier in time, not delay it.