Skip to content

Comment on The Way to the Wormhole

You are viewing one comment and its replies. View all comments · View the comment above

Comments

  • @MissingInteger@lemmy.zipOP· · edited

    I made it myself. Here is the ffmpeg command for anyone interested:

    ffmpeg -ss 00:22:47.700 -to 00:22:51.745 -i  *S05E14*  \
        	-filter_complex "[0:v][0:s:0]overlay" \
        	-an -c:v libsvtav1 -crf 34 -preset 0 -svtav1-params fast-decode=2  \
        	-g 300 -pix_fmt yuv420p -loop 0 \
    	'DS9 Way to the Wormhole (small)'.avif
    
    
    • -filter_complex "[0:v][0:s:0]overlay" hardcodes the dvd subs
    • -an -c:v libsvtav1 -crf 34 -preset 0 -svtav1-params fast-decode=2
      • no audio, svt-av1, fast-decode=2 for better playback (maybe just placebo)
    • -g 300 keyframe interval, should probably be higher since we don’t need any
    • loop 0 loops infinitely; for avif, in contrast to gif, you can actually specify how often it should loop.

    Funny thing, the low-quality webp thumbnail that lemmy generated is nearly twice the size of my encoded avif.
    236KiB webp thumbnail vs. 128KiB avif

    0
    • @SatyrSack@quokk.au· · edited

      I will try messing with something like that. My method for making AVIF animations is to pipe it to avifenc, because I couldn’t get ffmpeg to do it all:

      ffmpeg -hide_banner -loglevel error -i video.mp4 -pix_fmt yuv420p -f yuv4mpegpipe -fps_mode passthrough -vf "fps=25, scale=-1:480" -quality 80 - | avifenc --stdin image.avif >/dev/null
      

      EDIT: I just tried your ffmpeg-only command, and it worked, it was just a lot slower than my avifenc-piped command. Now that I think about it, I believe I had gotten it to work entirely with ffmpeg before, but decided to stick with my avifenc method simply because it is much more efficient.

      1
      • Change -preset 0 to a higher value; this is the lowest setting and it should in most cases severely slowdown encoding. A value of -preset 6 should be way faster. According to the ffmpeg wiki you can choose between 0 and 13 (13 only for debugging). Higher value → faster encoding. A lower value should give, in most cases, a smaller file (sometimes not; I had cases where 4 was smaller than 0).
        I only used 0 here since it was such a small clip of low-ish resolution and encoding with -preset 0 and -preset 4 were both nearly instantaneous. (size comparison here -preset 0 128K and preset 4 135K)

        0
        • Nice, changing that to 6 sure made it convert just as quickly as I expect. The resulting image quality seems fine as well, while actually being a slightly smaller file than what I get with my avifenc method.

          Any idea why it dumps out this extra info every time I run it? I specifically use -hide_banner -loglevel error to avoid this type of thing. Even putting >/dev/null at the end of the command doesn’t stop it.

          $ ffmpeg -y -hide_banner -loglevel error -i *.mp4 -an -c:v libsvtav1 -crf 34 -preset 6 -svtav1-params fast-decode=2 -g 300 -pix_fmt yuv420p -vf "fps=25, scale=-1:480" -quality 80 'test3'.avif >/dev/null
          Svt[info]: -------------------------------------------
          Svt[info]: SVT [version]:       SVT-AV1 Encoder Lib v3.1.2
          Svt[info]: SVT [build]  :       GCC 16.0.1 20260115 (Red Hat 16.0.1-0)   64 bit
          Svt[info]: -------------------------------------------
          Svt[info]: Level of Parallelism: 5
          Svt[info]: Number of PPCS 76
          Svt[info]: [asm level on system : up to avx512icl]
          Svt[info]: [asm level selected : up to avx512icl]
          Svt[info]: -------------------------------------------
          Svt[info]: SVT [config]: main profile   tier (auto)     level (auto)
          Svt[info]: SVT [config]: width / height / fps numerator / fps denominator               : 640 / 480 / 25 / 1
          Svt[info]: SVT [config]: bit-depth / color format                                       : 8 / YUV420
          Svt[info]: SVT [config]: preset / tune / pred struct                                    : 6 / PSNR / random access
          Svt[info]: SVT [config]: gop size / mini-gop size / key-frame type                      : 300 / 16 / key frame
          Svt[info]: SVT [config]: BRC mode / rate factor                                         : CRF / 34 
          Svt[info]: SVT [config]: AQ mode / Variance Boost                                       : 2 / 0
          Svt[info]: SVT [config]: sharpness / luminance-based QP bias                            : 0 / 0
          Svt[info]: -------------------------------------------
          
          1
          • Glad that it worked for you!

            I have no detailed knowledge of how to suppress specific information output of ffmpeg. I actually like that it vomits out so much information; I have my terminal scroll back set to essentially infinite and can so better analyze what went wrong or not quite right.


            Okay, now I know more: svt-av1 apparently pipes all its info as stderr, and that’s why -loglevel error doesn’t suppress it and why >/dev/null also doesn’t suppress it since this only redirects stdout to the void. You could do 2> /dev/null and suppress all stderr, or you could append it to a temp file with 2>> /tmp/ffmpeg-error.txt.

            But I found a better solution in a superuser answer:
            You can use the SVT_LOG environment variable like so:

            SVT_LOG=1 ffmpeg -y -hide_banner -loglevel error -i *.mp4 -an -c:v libsvtav1 -crf 34 -preset 6 -svtav1-params fast-decode=2 -g 300 -pix_fmt yuv420p -vf "fps=25, scale=-1:480" 'test4'.avif
            

            SVT_LOG=1 only gives actual errors, and you can make it permanent by adding export SVT_LOG=1 to your shell profile.


            Another thing: -quality 80 doesn’t do anything, and ffmpeg even gives a warning, which you suppressed. Quality is being controlled by -crf 34, see the wiki.

            0
            • Thanks! I will work that into my workflow.

              For the sake of comparison, I got JPEG-XL animations working from FFMPEG. But adjusting the values in the distance and effort options in the command below, I cannot get a file that is comparable to what I can get out of AVIF. Sure, it’s much better than GIF and pretty close to WebP, but compared to AVIF the JPEG-XL file is always either larger in file size or lower quality.

              ffmpeg -i test.mp4 -c:v libjxl_anim -f image2pipe -distance 8 -effort 5 -loop 0 -vf "fps=25, scale=-1:480" test.jxl
              
              1
      • I messed around a bit with avifenc and it gave me generally larger encodes with worse quality than just svt-av1 and rav1e with just ffmpeg. Also I remember getting numerous errors with ffmpeg complaining about piping this or that format. I wouldn’t even know which commands in my shell history actually worked. I have to try again with avifenc and experiment more with rav1e too both in ffmpeg and avifenc.

        0

Sign in to comment. Sign in