# mpv.conf

``````````````````````````````````````````````````````````````````
profile=sw-fast
vo=sdl
hwdec=no
scale=ewa_lanczossharp
cscale=ewa_lanczossharp

fps=29.970

# Low-latency
untimed
framedrop=no
#fps=60
no-demuxer-thread
vd-lavc-threads=1
video-sync=audio
interpolation=no
cache-pause=no
demuxer-lavf-o-add=fflags=+nobuffer
demuxer-lavf-probe-info=nostreams
demuxer-lavf-analyzeduration=0.1
video-latency-hacks=yes

fs
volume-max=250
sub-visibility=no
#audio-channels=1

screenshot-template='%F%wM%wS'

[extension.webp]
loop
[extension.jpg]
loop
[extension.jpeg]
loop
[extension.png]
loop
[extension.gif]
loop

[extension.mp3]
no-video
[extension.opus]
no-video
[extension.m4a]
no-video
[extension.flac]
no-video
``````````````````````````````````````````````````````````````````

The main purpose is to reduce latency, and to use only CPU. Reduci
ng latency really makes a difference. This is not about mpv, but t
urning audio files into opus with "-frame_duration 2.5" in ffmpeg
really helps you. Even without that option, opus has very low late
ncy. I use only opus, except for archiving. For CPU vs GPU, GPU is
harmful. Its process is hidden, and it makes things complex. Think
about GPU supporting problems in *nix.

The options below "#Low-latency" contain all the options in "profi
le=low-latency", except "audio-buffer=0". The exception is because
it desynchronizes between a video and an audio. The manual says "s
peed=1.01" is good, but it also desynchronizes.

"profile=sw-fast" is very helpful for software decoding. For fps,
I use 30000/1001 = 29.970. It is the fps of TempleOS. I also tried
"audio-channels=1", mono audio. But I realized that I'm not on tha
t level. For scales, ewa_lanczos* are the best. I'm only using ewa
_lanczossharp, because I like sharp images. For vo, SDL is the bes
t option for OpenBSD, because I don't use "vo=gpu". "vo=x11" tears
.