I need to count the number of frames in a video captured by a camera on a per-second basis. I haven't found a solution using ffmpeg or ffprobe (or something else) to output the number of frames per second (maintaining a constant frame rate is not guaranteed because of the capture mechanism and needs to be verified).
So far, I've needed to run ffmpeg and ffprobe separately. First, I run ffmpeg to trim the video:
ffmpeg -ss 00:00:00 -to <desired time in seconds> -i <in_video> -c copy <out_video>
Then, I run ffprobe to count the number of frames in the snippet:
ffprobe -v error -select_streams v:0 -count_frames -show_entries stream=nb_read_frames -print_format csv <out_video>
Is there one command to output the number of frames for each second in the video?
Run
ffmpeg -report -i <in_video> -an -vf select='if(eq(n,0),1,floor(t)-floor(prev_selected_t))' -f null -
In the generated report, search for select:1.000000
that will get you lines of the form
[Parsed_select_0 # 000001f413152540] n:270.000000 pts:138240.000000 t:9.000000 key:0 interlace_type:P pict_type:P scene:nan -> select:1.000000 select_out:0
The t is the timestamp and the n is the frame index. Check the frame index for each successive t. The difference is the frame count for that 1 second interval.
Related
I am trying to use the ffmpeg library to take two FLAC files and replace the audio in File A with the audio in File B at a given timestamp.
For example if File B was to be played at 00:02 and was a second long, playing the output it would be (00:00-0:01) File A Audio -> (00:02-0:03) File B Audio -> (00:04-...) File A Audio
To do this, I have tried the following
ffmpeg -y -i original.flac -i replacement.flac -acodec copy -ss 2 -to 3 -write_xing 0 result.flac
But this only produces the original audio between the specified timestamps.
Is there any way to achieve this within ffmpeg?
The typical method to do this would be the concat demuxer, but there are issues with FLAC extraction with duration header in the output, so you can use
ffmpeg -y -i original.flac -i replacement.flac \
-filter_complex "[0]atrim=0:2[Apre];[0]atrim=5,asetpts=PTS-STARTPTS[Apost];\
[Apre][1][Apost]concat=n=3:v=0:a=1" out.flac
Where 2 is the insertion point in seconds, and 5 is the insertion point + B's duration.
I need to merge audio and video using ffmpeg so that, it should result in a video with the same duration as of audio.
I have tried 2 commands for that requirement in my linux terminal. Both the commands work for a few of the input videos; but for some other input_videos, they produce output same as the input video, the audio doesn't get merged.
The commands, I have tried are -
ffmpeg -i wonders.mp4 -i Carefull.mp3 -c copy testvid.mp4
and
ffmpeg -i wonders.mp4 -i Carefull.mp3 -strict -2 testvid.mp4
and
ffmpeg -i video.mp4 -i audio.wav -c:v copy -c:a aac -strict
experimental output.mp4
and these are my input videos -
samplevid.mp4
https://vid.me/z44E
duration - 28 seconds
size - 1.1 MB
status - working
And
wonders.mp4
https://vid.me/gyyB
duration - 97 seconds
size - 96 MB
status - not working
I have observed that the large size (more than 2MB) of the input video is probably the issue.
But, still I want the fix.
I want to capture the RTSP stream from some IP cameras, and after looking around I found 2 great tools to do this: avconv and openRTSP
openRTSP -u user password rtsp://10.48.34.125/axis-media/media.amp
avconv -i "rtsp://user:password#10.48.34.125/axis-media/media.amp" -vcodec copy -f mp4 10.48.34.125.mp4
but for some voodoo reason when I need to use URLs without an specific extension, such as:
rtsp://user:password#10.48.34.46/
avconv returns 401 Unauthorized
so I'm stuck with openRTSP at the moment...
The thing is, unlike avconv, openRTSP outputs a raw file which is encoded to 25fps, which made some of my videos look like they where in fast-forward. I found a (cpu expensive) way to re-encode the file to a closer frame rate to what I need:
avconv -r 7 -i video-H264-1 -r 24 -f mp4 10.48.34.28.mp4
(in this example I'm forcing the frame rate of the raw file to be 7, and the frame rate of the output file to be 24. I tried using openRTSP build-in flags, but the output file still had a frame rate of 25: openRTSP -f 7 -u user password rtsp://10.48.34.145/mpeg4/media.3gp)
Sadly the video looks odd at certain points, and that's because the original stream sometimes has a variable frame rate (for example at night).
My question is, is there some way to deactive this default encondig to 25fps?
And why 25? I mean, isn't the norm 24?
try:
avconv -rtsp_transport tcp -i rtsp://server -an -vcodec copy -f mp4 10.48.34.28.mp4
if you want to change original video rate to 24 you must transcode it:
avconv -rtsp_transport tcp -i rtsp://server -an -vcodec libx264 -r 24 -f mp4 10.48.34.28.mp4
I can create slideshow with below command
ffmpeg -y -f image2 -r 1/15 -i image%d.jpeg -y -r 45 video.mp4
Video is created successfully ,, but its ignoring the 1st image..
for example in current folder i have
image1.jpeg
image2.jpeg
image3.jpeg
but slide show is created with image2.jpeg and image3.jpeg only..(i.e 30 second video with 2 image is created..but i expected 45 second video with 3 images displaying 15 sec per image)
What's wrong with your command is the -r 45 as an output option. You're telling ffmpeg to produce a video with 45 fps output rate. You probably don't want that.
Use the fps filter to set the framerate:
ffmpeg -r 1/15 -i img%d.jpeg -vf fps=25 output.mp4
If you use
ffmpeg -r 1/15 -i img%d.jpeg output.mp4
you will get a video with ~0.07 fps, and three frames, meaning a roughly 45 second long video (3 / 0.07). This gives the expected result, but it might not be playable in all players.
Note:
-y is not needed twice. Only supply it once before -i.
-f image2 is not needed.
Using -r as an output option to force another frame rate skips the first frame immediately in some players. I consider this a bug, which is now reported in this ticket .
This is definitely a strange question but I'm looking for a way to split an mp3 mix of 60 minutes into 60 separate 1 minute long wav files to use with an audio fingerprinting API like Echonest.
Is this possible in a single ffmpeg command or would I have to run multiple iterations of ffmpeg with a the following values:
-ss is the startpoint in seconds.
-t is the duration in seconds.
You can use the segment muxer in ffmpeg:
ffmpeg -i input.mp3 -codec copy -map 0 -f segment -segment_time 60 output%03d.mp3
For a 4 minute input this results in:
$ ls -m1 output*.mp3
output000.mp3
output001.mp3
output002.mp3
output003.mp3
Since -codec copy enables stream copy mode re-encoding will be avoided. See the segment documentation for more information and examples.