Search options (Page 2 of 9)

If dawkinscm's config doesn't give good enough results, and tinkering with the profile settings might make it slower, i would also test with the default automatic profile + masking disabled (to confirm if it's better/worse).
And lowering the monitor resolution a bit if it's already at 3840x2160 (but you only watch movies at 1920x1080).
You can also use the "Alter video frame size" button to lower the amount of pixels that get interpolated:
https://gyazo.com/92109d19b50e5eeb29a72123fc0ea5a8
But the quality loss is not worth it if u ask me.

Surely applying a combination of above will get u to stable 120 fps.

Yes of course. All our hardware setups are different.
Last ideea: selecting "Film" here requires much more processing power:
https://gyazo.com/8ca66edc6222cea98e6b79436b0363d9
So even if you only watch live-action movies, i would still recommend the "Animation" button for less memory usage. (Or the old 4.4-4.6 RIFE v2 models cause they eat less)

They released new mpv version:
https://github.com/mpv-player/mpv/releases/tag/v0.41.0

Good job!

The higher frame rate you go, the artifacts become less noticeable since they stay on screen less time.
According to "Artifacts masking" description:
https://gyazo.com/abc2d875f0bfaf071249dd4ca0546283
Enabling it will slow down smoothness and blur everything (so you won't notice movement artifacts).
But if you're system CAN run already at 240 fps and above, you might as well disable all artifacts masking everywhere to maximize picture clarity and minimize blur.

If i set real_timestamp_tolerance to 1, it shows 24/1000 fps but panning scenes move choppy exactly like original 24 fps, meaning it moves like 1x with scaling effect ON.
https://gyazo.com/1e5904f7c6a47c63612a93f2f134247f
Where is the black insertion setting and what does it do?
For performance increase, just change HDR support checkbox to OFF, that's the real memory hog.

After further tweaking, i think these settings actually improve smoothness, giving stable fps:
https://gyazo.com/96617651c91f7e135f2b22eb65fa900d
real_timestamp_tolerance = 0
This sets LSFG to show only generated frames.
This is needed because setting it to 1 would show all real frames, disabling interpolation, meaning the default of 0.05 at 1000 fps would show you 50 static frames breaking frame-pacing.
So a value of 0 will correctly show 100% smooth frames.

Thanks Ricky!

https://github.com/hzwer/Practical-RIFE … 2798826664

Replace both mpv.exe files with latest version and try again?
(there's like 2 years difference between 0.35 and 0.39 so probably just updating it can fix it)
https://mpv.io/installation/

yes, MPC Video Renderer is a component of both MPC-HC and MPC-BE
usually latest version haz more functionalities
i found that link here:
https://github.com/Aleksoid1978/VideoRe … tly-builds

Or a SVP reddit community where people can post random questions exactly like on this forum?
example:
https://www.reddit.com/r/mpv/

Most common issue i saw it has is that it does not detect the app's correct FPS
So if u play a 24 fps video with 2x, it should say 24/48 if it detects correctly how many frames that app is running with.
If it says 144/288 means it detects your monitor hz instead. xD
Ways to fix it is to install it directly from steam, close some overlay apps that might interfere with it, or set this in Nvidia Control Panel: https://gyazo.com/63bf77412bcc01e6dcb246143514105d
Or disable the extra monitors if u're multi monitors setup.

If both oriento and https://www.svp-team.com/forum/viewtopic.php?id=7530 mentioned a sudden performance drop compared to previously
it could mean that one of the recent SVP updates like updating python version or one of the recent hotfixes increase the processing power required for live interpolation just above their system's capabilities
but we don't have a downgrade SVP button to let them investigate

best scenario would be if devs try to interpolate a video on one of their machines at high enough processing power that they can reproduce their issue, but not too high so rollbacking a couple of SVP updates would see a noticeable performance improvement

this means devs should go through "A few weeks ago..." of SVP changes in order to confirm if slowdown is related to a critical component that requires updates or not to conclude if it's fixable or if we should just blame Nvidia's latest drivers

offtopic question:
this looks promising and uses RIFE and Vapoursynth, would it be possible to integrate on top of SVP with a checkbox?
https://github.com/routineLife1/VS-DRBA
https://github.com/routineLife1/DRBA
or probably not worth it

i remember bumping into something similar months ago when i played with that menu
i suggest to select a configuration on that menu then restart SVP to apply it, probably it happens cause of the resizing that happens while the video is already running

average user just plays the video once so RIFE keeps same width/height from start to end of video

I also see that the RIFE developer added some extra parameters for the new TensorRT version, not sure if mandatory:
https://gyazo.com/87bdd69a3c5e6d3f934b003cbe4a71b1