вс, 12 нояб. 2023 г., 20:23 Aditya Dandavate <adityadandavat...@gmail.com>:
> I have an Intel Pentium G3240 processor. > This is a quite a low end processor and even x264 etc is cpu-intensive and > even encoding with this encoder is really slow. (6.0 to 6.5 fps with medium > preset). > > I encoded some videos with mpeg4 video encoder and h264_qsv encoder. The > results I got was the speed of h264_qsv with preset veryslow, is almost the > same to mpeg4 -q:v 5, with 15000k bitrate for both encoders. libopus was > audio encoder with 320kbps bitrate for both encoding settings. > (h264_qsv = 16-17 fps, mpeg4 = 18-19 fps). > well, isn't it depend on frame size and fps too? 15 Mbit/s * h264 sounds quite good bitrate for say, 1080p30, but h264 in general allows for less bitrate at similar quality. So watch them both and decide? I think at 25+ MB/s even old standby like mpeg2 can deliver not so bad quality at full hd? I used to think that empirically mpeg4 compress 1/3 better than mpeg2, then h264 shaves thid more, and so on, at progressively highr cpu usage on both decode and encode. Of course there is plenty of knobs to tune, vbr vs cbr vs two-pass vbr ... this rabbit hole goes deep! > So, Should I stick to h264_qsv or mpeg4 ? In general, which encoder gives > higher quality at the same bitrate ? > Hardware encoders used to be not good at 2-4-6 Mbs range for live streaming, but from existence of some bluray compat option in nvenc wrapper I can guess at BluRay bitrates they might be quite good? watch water/fire scene - usually they hard to encode. > _______________________________________________ > ffmpeg-user mailing list > ffmpeg-user@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-user > > To unsubscribe, visit link above, or email > ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe". > _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".