一个非常简单的方法是ffmpeg
在后台运行每个进程,然后等待它们完成,然后再继续crf
.
for crf in 10 15 20 23 25 30 35 ; do
for vid_preset in medium slow veryslow; do
ffmpeg -r 25 -i pics/pic_%04d.png -vcodec libx264 \
-crf $crf -vpre $vid_preset anim_crf${crf}_${vid_preset}.mp4 &
done
wait
done
如果你已经parallel
安装了,你可以用它来繁忙地运行所有 21 个进程。
# Run 4 jobs at a time, starting a new job whenever one completes.
parallel -j 4 ffmpeg -r 25 -i pics/pic_%04d.png \
-vcode libx264 -crf {1} -vpre {2} \
anim_crf{1}_{2}.mp4 ::: 10 15 20 23 25 30 35 ::: medium slow veryslow
{1}
被来自第一个输入源的选择(在第一个输入源之后:::
)替换为来自{2}
第二个输入源的选择。一个例子更简单地解释它:
$ parallel echo {1} {2} ::: a b c :: 1 2
a 2
a 1
b 1
b 2
c 1
c 2
bash
4.3 或更高版本中的混乱方法(wait -n
引入时):
max_jobs=4
running=()
for c in 10 15 20 23 25 30 35; do
for v in medium slow veryslow; do
ffmpg ... &
running+=($!)
# If we're at capacity, wait for a job to complete
if (( ${#running[@]} == $max_jobs )); then
wait -n
# Any where from 0 to max_jobs - 1 could still be running
running=( $(jobs -p) )
fi
done
done