Installation:
mpv.io
Wiki:
github.com
Manual:
Stable: mpv.io
Git: mpv.io
User Scripts:
github.com
High quality video output profile (goes into mpv.conf):
profile=gpu-hq
Configuration Files:
mpv.io
mpv.io
Input.conf:
github.com
Post your system specs and config if you're asking performance related questions.
Windows Builds:
Stable: mpv.srsfckn.biz
Git: sourceforge.net
Evaluating mpv's upscaling algorithms:
artoriuz.github.io
Other urls found in this thread:
mpv.io
sourceforge.net
diff.pics
diff.pics
diff.pics
diff.pics
gist.github.com
madvr.com
twitter.com
>Comfy OP.
mpvnet rocks you elitists faggots
can i change default volume to 75 ? So When mpv start it always back to 75
Does somebody use mpv on Android?
My desktop config file doesn't work on Android.
Are the options the same or why it doesn't work?
Stats still hasn't been fixed while using gpu-api=vulkan?
wm4 not on the boat ofc they haven't
>80 open pull requests
JEEB
So how do I install this shit? The wiki isn't very helpful. Do I need to choose a "build" and download that? Which build should I use when starting out?
After a 1000 confs, I've decided this is the most resource efficient but also without sacrificing too much visual quality conf:
# Scaling
scale=ewa_lanczossharp
cscale=ewa_robidouxsharp
dscale=ewa_robidouxsharp
scaler-lut-size=10
# Shaders
glsl-shader="C:\Program Files\mpv\shaders\SSimSuperRes.glsl" #With robidouxsharp parameters
glsl-shader="C:\Program Files\mpv\shaders\LumaSharpenHook.glsl" #Strength at 0.35 with pyramid shaped pattern.
I have 3ms subtimings while upscaling a 720p video to 1440p, pretty good.
RX 480 8GB
Ryzen 7 [email protected]
No hardware decoding.
Are u using windows or linux?
Using Windows 7 btw
...
Download the latest shinchiro build from here:
sourceforge.net
Inside the "installer" folder there's a file called "install.bat", execute it with administrator rights and you're done, you've installed mpv, congratulations.
Is this better than lachs0r's build?
Is more up to date tho sometimes can be buggy
Thanks user
What is it doing instead of working? Note: Android OpenGL drivers are utter trash. Anything that uses the advanced rendering path will not work on Snapdragon/Adreno, so even if your phone is powerful, you probably won't be able to use a desktop config. Apart from that the options are the same.
I only got profile=gpu-hq to work without lagging after adding a line about the GPU context being Wayland. And then everything was beautiful.
Do nothing.
I can't even change subtitle color.
slang=eng,en,enUS,en-US,English
ass-style-override=force
sub-ass-force-margins
sub-text-font-size=68
sub-color='#ffd200'
>720p video to 1440p
Why not just use FSRCNNX?
Timings to up to 6ms with FSRCNNX_8
Besides SSimSuperRes with robidouxsharp parameters is almost as good.
anything below 16ms is fine though?
>robidouxsharp
Don't make me fall for memes
Not if you're using a 144hz display.
That's stupid, you're not gaining visual fidelity by watching at 144Hz, but you would by using FSRCNNX.
I'm gaining judder free 24hz playblack tho, when combined with video-sync=display-resample that is.
It's not a meme user, it's actually pretty good:
diff.pics
diff.pics
diff.pics
diff.pics
The FSRCNNX used in the comparisons is FSRCNNX_16.
>I'm gaining judder free 24hz playblack tho, when combined with video-sync=display-resample that is.
You don't need 144Hz for that, 72 is enough and should be possible on a decent monitor.
Aren't you just repeating the same frame six times when watching 24fps content on a 144hz display? What's the point of render times that low?
What font do you use for subtitles? It's very annoying to test them out manually, I have like 500 fonts. Is there a website with font screenshots (simulating a video, position and size matters)?
>cscale=ewa_robidouxsharp
trash
># Shaders
>glsl-shader="C:\Program Files\mpv\shaders\SSimSuperRes.glsl" #With robidouxsharp parameters
>glsl-shader="C:\Program Files\mpv\shaders\LumaSharpenHook.glsl" #Strength at 0.35 with pyramid shaped pattern.
share the parameters
and what scaler-lut-size placebo even means?
>dscale=ewa_robidouxsharp
and this is absolute garbage
It's perfectly explained in the manual.
I've seen your comparisons in a previous thread, but I still don't know how you managed to make ewa_lanczossharp + SSSR (with RobidouxSharp params) look this good.
Nope and nope, I test my shit like a 1000 times, dscale=mitchell is too soft, dscale=ewa_robidouxsharp provides the perfect sharpness without introducing ringing, cscale=lanczos and cscale=ewa_lanczossharp cause ringing in the chroma channels, cscale=haasnsoft is too soft, cscale=ewa_robidouxsharp provides a sharper image than haasnsoft without introducing ringing.
You need to change the 31 and 79 to pic related:
0.2620145123990142 and 0.3689927438004929
See the comments here for more info:
gist.github.com
In LumaSharpenHook just change the line 21 to 0.35 and the line 26 to 4
See
*You need to change the 31 and 79 line
now test it 1001st time on this image madvr.com
How am I suppose to upscale an image so big?
I'm talking about factual proofs, not placebo theory.
Wow...
>madvr
YIKES!
Wow what?
But what the manual says about it is completely factual. There's nothing more than that.
I know how you changed the SSSR parameters. I'm just not able to achieve as good results as you.
Thx user
I don't know what to tell you user.
It worked for me, I don't know why isn't working for you.
why did igv get rid of FSRCNNX x3 and x4
He's telling you to stop using a toaster.
shouldn't direct upscaling be more consistent than two step?
what's your interpolation settings
Oversample. Everything else is too blurry. madvr only has oversample because of this reason.
Which one?
I tend to agree
tscale I guess, I don't really know what the others do
I prefer oversample. I've tried a couple of the meme options that's been cropping up, and they're either too blurry for my tastes, or look just like oversample.