Vera S05 Libvpx — Best =link=

-c:v libvpx -cpu-used 4 -crf 10 -b:v 1M VP8 at crf 10 looks excellent and runs 3x faster than VP9 on the same chip. Leverage the S05’s CedarX decoder for input. Use:

export FFMPEG_THREADS=4 export GOMP_CPU_AFFINITY=0-3 Then force -bufsize to a low value (e.g., -bufsize 2000k ) to prevent OOM (Out of Memory) crashes. For older Vera S05 kernels, libvpx (VP8) is significantly faster. If VP9 is too heavy, fall back to VP8 with: vera s05 libvpx best

ffmpeg -i input.mkv -c:v libvpx-vp9 \ -cpu-used 2 \ -crf 30 \ -b:v 900k \ -maxrate 1800k -bufsize 3600k \ -threads 4 -row-mt 1 -tile-columns 2 \ -g 120 -tile-rows 0 \ -pass 1 -f webm /dev/null -c:v libvpx -cpu-used 4 -crf 10 -b:v 1M

./configure --enable-neon --enable-runtime-cpu-detect The generic binaries shipped with most Linux distros leave 30-40% of the Vera S05’s potential on the table. Unlock it, and you'll truly have the this little board can offer. For older Vera S05 kernels, libvpx (VP8) is

-hwaccel drm -hwaccel_output_format drm_prime This frees CPU cycles for the software libvpx encoder. To verify your settings, use ffmpeg ’s built-in filters:

ffmpeg -i input.mkv -c:v libvpx-vp9 \ -cpu-used 0 \ -crf 18 \ -b:v 0 \ -threads 4 -row-mt 1 -tile-columns 2 \ -speed 0 \ -lag-in-frames 25 \ -auto-alt-ref 1 -sharpness 0 \ -pass 1 -f webm /dev/null ffmpeg -i input.mkv -c:v libvpx-vp9 -cpu-used 0 -crf 18 -b:v 0 -threads 4 -row-mt 1 -tile-columns 2 -speed 0 -lag-in-frames 25 -auto-alt-ref 1 -sharpness 0 -pass 2 -f webm output.webm