appwiki:ffmpeg [ShiningWiki]

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== FFmpeg - the core command app of open source transcoder ======

  * free and available as command line software for windows, mac, linux
  * download: http://ffmpeg.org/
  * input video format: 3GP, AVI, FLV, Matroska, MP4, MPEG-PS, Ogg, QuickTime, SVCD, TS, TOD, VCD, WMV
  * output video format: 3GP, AVI, Blu-ray, DVD, FLV, MKV, MP4, MPEG-PS, Ogg, mov, super VCD, TS, VCD, Webm, WMV

  * Doc: https://ffmpeg.org/

====== Common Convertion ======

  * disable ffmpeg feedback by adding in 1st flag <code bash>ffmpeg -hide_banner -loglevel panic -i test.jpg test.png</code>

  * MTS 50i interlaced video to progress video 50p (ref: [[http://blog.tahvok.com/2013/10/deinterlacing-and-converting-mts-video.html|link]]) <code bash>ffmpeg -i in.MTS -vf yadif=1 -acodec ac3 -ab 192k -vcodec mpeg4 -f mp4 -y -qscale 0 out.mp4</code>

  * image sequence to mp4 in cmd (25fps, FHD size, pic.0001.png naming, H264, quality 15-25)<code bash>ffmpeg -r 25 -f image2 -s 1920x1080 -i pic.%04d.png -vcodec libx264 -crf 25  -pix_fmt yuv420p test.mp4</code>

  * **image sequence to MOV** with start frame detection (e.g h264 mov, prores 422)
    * h264 mov <code python>
import os,glob,shlex,subprocess
imgSeqPath = 'D:/ExamplePathRoot/ExampleName/ExampleVersion/example.%04d.exr' # 0001 format image names
movPath = 'D:/OutputMov/ExampleName/example.mov'

file_list = glob.glob( os.path.join(os.path.dirname(imgSeqPath), re.sub('\.?%0\d+d$', '*', base) + ext ))
if len(file_list) == 0:
    return
prefix, ext = os.path.splitext(os.path.basename(file_list[0]))
start_number = re.findall('\d+$', prefix)[0]
enc = 'D:/toolPath/ffmpeg.exe -y -r 24 -start_number {0} -i "{1}" -an -vcodec libx264 -preset slow -crf 22 -threads 0 "{2}"'.format(start_number, imgSeqPath, movPath)
subprocess.Popen(shlex.split(enc), stdout=subprocess.PIPE, stderr=subprocess.PIPE)
</code>
    * proRes422 <code python>
# profile: 0 proxy, 1 LT, 2 SQ, 3 HQ
# pix format: yuv422p10le , for 444, use yuva444p10le
# -qscale:v 11: best 0-32 worst
# ref: https://trac.ffmpeg.org/wiki/Encode/VFX
enc = 'D:/toolPath/ffmpeg.exe -y -r 24 -start_number %s -i "%s" -an -c:v prores_ks -profile:v 2 -pix_fmt yuv422p10le -vendor ap10 "%s"' % (start_number, imgSeqPath, movPath)
</code>

  * image conversion and resize (width 320px) <code>
ffmpeg -i input.jpg -vf scale=320:-1 output_320.png
</code>

  * convert mov with alpha to webm with alpha <code>
ffmpeg.exe -i transparent_video_audio.mov -c:v libvpx-vp9 -pix_fmt yuva420p -b:v 2000k out.webm
</code>

  * convert mpg to mov h264 <code>
ffmpeg.exe -i "d:\out\me.mpg" -c:v libx264 -c:a libfaac -crf 20 -preset:v veryslow "d:\out\me_out.mov"
</code>

  * copy portion of video <code>
# about -ss before -i and after -i  difference, which can make cut video works or not works
# ref: https://ottverse.com/trim-cut-video-using-start-endtime-reencoding-ffmpeg/

ffmpeg.exe -i video.mp4 -c copy -ss 00:00:09 -to 00:00:21 video_cut.mp4

# -ss in front means: start time (from 9s), duration time (21s)
ffmpeg.exe -ss 00:00:09 -i video.mp4 -c copy -to 00:00:21 video_cut.mp4
</code>
  * combine list of videos into 1 video
    * ref: https://stackoverflow.com/questions/7333232/how-to-concatenate-two-mp4-files-using-ffmpeg

  * gif maker <code>
ffmpeg -i input.mp4 -vf "fps=10,scale=640:-1:flags=lanczos,split[s0][s1];[s0]palettegen[p];[s1][p]paletteuse" -loop 0 output.gif
</code>

  * 2x2 stack <code>
#ref: https://ottverse.com/stack-videos-horizontally-vertically-grid-with-ffmpeg/
ffmpeg -i input0.mp4 -i input1.mp4 -i input2.mp4 -i input3.mp4 -filter_complex "[0:v][1:v]hstack=inputs=2[top]; [2:v][3:v]hstack=inputs=2[bottom]; [top][bottom]vstack=inputs=2[v]" -map "[v]" finalOutput.mp4
</code>

  * image to square size (like convert image to texture size format) <code>
ffmpeg -i image.jpg -vf "scale=1024:1024:force_original_aspect_ratio=decrease,pad=1024:1024:-1:-1:color=white" -q:v 1 tex_image.jpg

# longest side, so basically original size with padding to square
ffmpeg -i image.jpg -vf "scale=max(iw\,ih):max(iw\,ih):force_original_aspect_ratio=decrease,pad=max(iw\,ih):max(iw\,ih):-1:-1:color=white" -q:v 1 tex_image.jpg
</code>

====== GIF related ======

  * use every 1 from out of 10 frame from a huge frame-by-frame gif recording, and save as a 600px width gif <code dos>
ffmpeg.exe -i big_recording.gif -vf "select=not(mod(n-1\,10)),scale=600:-1" -vsync vfr light_10th_600w.gif
</code>