In my last post I mentionned how to use the GPU to re-encode video. A friend mentionned that there is a tradeof regarding file size. GPU encoding is much faster, but produces files that are also much bigger, around twice as big as the cpu encoded one.
I ran a quick test on my machine, using this video, size: 55.2MiB, originally encoded with VP9.
The main command is roughly:
ffmpeg -y -loglevel repeat+info -i original.webm -c:v av1 av1_cpu.mkv to get a cpu transcoding
or
ffmpeg -y -loglevel repeat+info -vaapi_device /dev/dri/renderD128 -i original.webm -vf 'format=nv12|vaapi,hwupload' -c:v av1_vaapi av1_gpu.mkv
for gpu transcoding.
| video codec | gpu size | cpu size | gpu time | cpu time |
------------------------------------------------------------
| av1 | 145 MiB | X | 7s | (1) |
| H.265 | 55 MiB | 24 MiB | 7s | 44s |
| H.264 | 109 MiB | 54 MiB | 9s | 21s |
| VP9 | (2) | 38.7 MiB | (2) | 3m37s |
Notes:
- I didn't complete that. ffmpeg was indicating a speed of around 0.05x, so the original 3:50 would take an estimated hour and half and I have better things to do.
- my gpu doesn't support gpu encoding for VP9. I still wanted to see the file size if I re-encoded it.
So that's one thing to keep in mind, whether you value speed of transcoding or file size.