26

(55 replies, posted in Using SVP)

dawkinscm wrote:
RickyAstle98 wrote:
dawkinscm wrote:

Dunno. I read about it somewhere that flip gets similar performance to exclusive that's why I re-enabled Nvidia MPO, but if you turn on exclusive then it is smart enough to bypass itself or something like that.

The MPO doesnt work correctly with Windows 10 system, thats why LSFG only allows DXGI renderer, DWM>Composed Flip instead of FSE>Hardware Composed Flip (your behavior)

The SpecialK somehow solved all judder/jerks/stutter problems, continuously holds refresh rate much agressively, but unstable on seek, I dont want to lose performance/smoothness in games for video players, SpecialK helps, trying to inject SpecialK window management through mpv library!

UPD: MPV Interpolation argument yes and D3D11 sync intervals solved the problems, but unstable enough!

Never heard of Special K but it looks interesting. Some of what it does is kind of what I'm doing but with Windows 11.

BTW what do you do with the Sync intervals, disable it?

Update:I've been testing my new settings with fast action. But I just tested my new settings with slow panning shots and it forced me to make some changes to my original recommendations.

As of today my settings are SVP: Rife v4.25, IC 8%, Blend Adjacent Frames. MPV: Interpolation on, resample audio. GPU: 577 or 566.36. Windows 11 Optimization for Windowed Apps: MPV

Optical flow seems to also help possibly removing the need for the overclock but at the cost of the occasional major stutter so nope.

The sync intervals prevents NVIDIA default driver compositor deliver frames non-directly, monitor VRR in charge, but breaks at seek again!
As I said, this is just VRR framepacing issue through the driver level, older drivers or LSFG ontop prevents monitor spontaneous refreshes, as you can see, earlier, I noticed how much VRR jumps got through standard MPV playback on new stable drivers, which not happens with LSFG ontop or after SpecialK injection, SpecialK makes 72FPS content not only smooth, but fast, not sure how that effect occurs, tested Gemini movie at real 60FPS compared to 24>60 and was surprised how SpecialK deliver frames, but I dont know how to inject SpecialK through MPV swapchain directly!

UPD: Found best solution to change present methods through NVPI program (inspector)! WORKS!!!

27

(55 replies, posted in Using SVP)

dawkinscm wrote:
RickyAstle98 wrote:
dawkinscm wrote:

Cool. We can only use what works.  Since I'm using Exclusive Fullscreen it's probably bypassing Flip and using DWM anyway lol.

Nope, doesnt happen to me, Exclusive Fullscreen behavior the same as before, destroyed framepacing with/without, thats your Win11 feature, my Win10 cant bypass that maybe?

Dunno. I read about it somewhere that flip gets similar performance to exclusive that's why I re-enabled Nvidia MPO, but if you turn on exclusive then it is smart enough to bypass itself or something like that.

The MPO doesnt work correctly with Windows 10 system, thats why LSFG only allows DXGI renderer, DWM>Composed Flip instead of FSE>Hardware Composed Flip (your behavior)

The SpecialK somehow solved all judder/jerks/stutter problems, continuously holds refresh rate much agressively, but unstable on seek, I dont want to lose performance/smoothness in games for video players, SpecialK helps, trying to inject SpecialK window management through mpv library!

UPD: MPV Interpolation argument yes and D3D11 sync intervals solved the problems, but unstable enough!

28

(55 replies, posted in Using SVP)

dawkinscm wrote:
RickyAstle98 wrote:
dawkinscm wrote:

I would be surprised if LFSG ignores filp models since modern Nvidia drivers all have Multiplane Overlay built in specifically to work with it. This was one of the Nvidia features I turned off not knowing that I was probably hurting more than helping.
I wouldn't play with the MPV swapchain. From my testing I found that the default is ideal so that's what my SVP Nvidia settings also use. I forgot to add this to my settings list because I set this up some time ago.
As for transcoding speed, I only look at the end results. Are packets dropped? Is the GPU stuck on 100%? Is SVP stable around 0.99 to 1.01 or is it fluctuating? My current answers are no, no and stable.

I think ignores because of this magic things, running LSFG at same target FPS as RIFE - making mpv looks smooth, very smooth, despite not generating frames at all!
Also I think LSFG ignores newest compositors because how quickly my monitor switches and holds refresh rate, currently at stable 581 (yes reverted back for games), during standard mpv playback without any interpolation, VRR fluctuating!

UPD: Running external video capture to get VRR jumps via SpecialK measurement, mpv during 1 minute scene at FIXED 24FPS fluctuates 210 times (no interpolation), and only 30 with LSFG ontop, DWM works flawlessly!

Cool. We can only use what works.  Since I'm using Exclusive Fullscreen it's probably bypassing Flip and using DWM anyway lol.

Nope, doesnt happen to me, Exclusive Fullscreen behavior the same as before, destroyed framepacing with/without, thats your Win11 feature, my Win10 cant bypass that maybe?

29

(55 replies, posted in Using SVP)

dawkinscm wrote:
RickyAstle98 wrote:
dawkinscm wrote:

Thanks for the the heads up smile

Update: I watched a full 7 minute sequence with zero stutters and is surprisingly smooth, considering IC 6% but 8% is more reliably stutter free.

As of today my settings are SVP: Rife v4.25, IC 8%, Blend Adjacent Frames. MPV: Interpolation on, resample audio. GPU: 590.26, Smooth Motion (Globally) On. Windows 11: Optimizations for Windowed Games.

These SVP settings fixes Rife's double image issues with fast vertical motion in movies like Dr Strange 2. Also GPU usage is greatly reduced. Maybe this is because it doesn't try to interpolate as many frames?.
These MPV settings stops it from dropping frames and helps with any remaining smoothness issues after SVP interpolation. I also use exclusive full screen but for most users I don't think this is "needed".
These Nvidia settings are not the only option but these work as well as any. A mild overclock still works best which is strange considering the GPU usage but maybe the peaks are higher?
The Windows settings are for MPV and other relevant software to work optimally with Windows 11.

I'm sure there's other ways of doing this, but this is the first time I have an explainable reason for each setting instead of just hope. This works for my VR setup so YMMV.

I will try to play with mpv swapchain parameters, LSFG ignores newest drivers flip models, producing smoothest image through DWM compositor, why not to force mpv to choose DWM composition too? Because before 566.14 driver, DWM was default!

Wait, 590.26 FIXED some issues?!

I would be surprised if LFSG ignores filp models since modern Nvidia drivers all have Multiplane Overlay built in specifically to work with it. This was one of the Nvidia features I turned off not knowing that I was probably hurting more than helping.
I wouldn't play with the MPV swapchain. From my testing I found that the default is ideal so that's what my SVP Nvidia settings also use. I forgot to add this to my settings list because I set this up some time ago.
As for transcoding speed, I only look at the end results. Are packets dropped? Is the GPU stuck on 100%? Is SVP stable around 0.99 to 1.01 or is it fluctuating? My current answers are no, no and stable.

I think ignores because of this magic things, running LSFG at same target FPS as RIFE - making mpv looks smooth, very smooth, despite not generating frames at all!
Also I think LSFG ignores newest compositors because how quickly my monitor switches and holds refresh rate, currently at stable 581 (yes reverted back for games), during standard mpv playback without any interpolation, VRR fluctuating!

UPD: Running external video capture to get VRR jumps via SpecialK measurement, mpv during 1 minute scene at FIXED 24FPS fluctuates 210 times (no interpolation), and only 30 with LSFG ontop, DWM works flawlessly!

Repeatable performance experiment (transcoding speed) (RTX 4070)

Driver 546.17 (output 2880 frames 1280x736 opt w/TRT 9.1.0)
60>120 ~304fps (4.4v2)
60>120 ~284fps (4.6v2)
60>120 ~260fps (4.7v2)
60>120 ~250fps (4.9v2)
60>120 ~210fps (4.12v2)
60>120 ~200fps (4.14v2)

Driver 565.90 (output 2880 frames 1280x736 opt w/TRT 10.13.0)
60>120 ~308fps (4.4v2)
60>120 ~286fps (4.6v2)
60>120 ~262fps (4.7v2)
60>120 ~252fps (4.9v2)

Driver 581.29 (output 2880 frames 1280x736 opt w/TRT 10.13.0)
60>120 ~292fps (4.4v2)
60>120 ~272fps (4.6v2)
60>120 ~248fps (4.7v2)
60>120 ~242fps (4.9v2)

HAGS: On

31

(55 replies, posted in Using SVP)

dawkinscm wrote:
RickyAstle98 wrote:
dawkinscm wrote:

That maybe because it's using "smooth motion generation". I got a hacked version of that technology kind of working with my 40 series card, but it will be great to see it in action properly working. Thanks!

It doesnt, disabled!

UPD: Some people said that dev driver 590.10 also better with competitive games, the quickest driver by latency, need to test with same parameters!

Thanks for the the heads up smile

Update: I watched a full 7 minute sequence with zero stutters and is surprisingly smooth, considering IC 6% but 8% is more reliably stutter free.

As of today my settings are SVP: Rife v4.25, IC 8%, Blend Adjacent Frames. MPV: Interpolation on, resample audio. GPU: 590.26, Smooth Motion (Globally) On. Windows 11: Optimizations for Windowed Games.

These SVP settings fixes Rife's double image issues with fast vertical motion in movies like Dr Strange 2. Also GPU usage is greatly reduced. Maybe this is because it doesn't try to interpolate as many frames?.
These MPV settings stops it from dropping frames and helps with any remaining smoothness issues after SVP interpolation. I also use exclusive full screen but for most users I don't think this is "needed".
These Nvidia settings are not the only option but these work as well as any. A mild overclock still works best which is strange considering the GPU usage but maybe the peaks are higher?
The Windows settings are for MPV and other relevant software to work optimally with Windows 11.

I'm sure there's other ways of doing this, but this is the first time I have an explainable reason for each setting instead of just hope. This works for my VR setup so YMMV.

I will try to play with mpv swapchain parameters, LSFG ignores newest drivers flip models, producing smoothest image through DWM compositor, why not to force mpv to choose DWM composition too? Because before 566.14 driver, DWM was default!

Wait, 590.26 FIXED some issues?!

dawkinscm wrote:
RickyAstle98 wrote:
dawkinscm wrote:

"MPV looks insane" Explain?

Under high GPU load, 565.90 causing stutters, but was buttery smooth all the time with average loads, 590.26 have no stutters under high load at all!
Noticed how quick my monitor VRR switches and holds, not happens with 566+ drivers!

That maybe because it's using "smooth motion generation". I got a hacked version of that technology kind of working with my 40 series card, but it will be great to see it in action properly working. Thanks!

It doesnt, disabled!

UPD: Some people said that dev driver 590.10 also better with competitive games, the quickest driver by latency, need to test with same parameters!

dawkinscm wrote:
RickyAstle98 wrote:
dawkinscm wrote:

Strange because 566.36 always worked well for me. 577.00 was maybe even better and the latest official Nvidia release is as good as the others. So as I said it will be a local setup issue. Doesn't mean that "you" are doing anything wrong. It just means that there's something in your setup that is the real cause. Could be a card limitation or needing a "clean install" not an express install. Could be Windows 10/11 settings. For me it was kind of all of the above plus more because I'm using VR.

The dev driver 590.26 breaks framepacing in games now, MPV looks insane, even better than 565.90 legacy now! Unstable, one game already crashed twice!

"MPV looks insane" Explain?

Under high GPU load, 565.90 causing stutters, but was buttery smooth all the time with average loads, 590.26 have no stutters under high load at all!
Noticed how quick my monitor VRR switches and holds, not happens with 566+ drivers!

dawkinscm wrote:
RickyAstle98 wrote:
dawkinscm wrote:

That's why I said "most driver issues are game specific". But I am using 581.42 with no issues so not sure what is happening there unless .29 has specific issues. I'm watching full 4k (3840x2160) 3D movies and running a remote desktop at the same time with my GPU regularly dipping well below 100% and no stutters so far.

No, 581.42 is the same as 29, different framepacing in different applications, which was the same for all applications for legacy 565.90 driver, 2 drivers before NVIDIA App transformation!

UPD: I will try dev drivers, 590.26 for test!

Strange because 566.36 always worked well for me. 577.00 was maybe even better and the latest official Nvidia release is as good as the others. So as I said it will be a local setup issue. Doesn't mean that "you" are doing anything wrong. It just means that there's something in your setup that is the real cause. Could be a card limitation or needing a "clean install" not an express install. Could be Windows 10/11 settings. For me it was kind of all of the above plus more because I'm using VR.

The dev driver 590.26 breaks framepacing in games now, MPV looks insane, even better than 565.90 legacy now! Unstable, one game already crashed twice!

dawkinscm wrote:
RickyAstle98 wrote:
dawkinscm wrote:

Told you lol. Most of the driver issues are either game specific or as you have found out, they are local setup issues.  BTW I'm now using the latest 58x driver with no issues. So it looks like in my case it was mostly local setup issues.

But how different drivers breaks framepacing in different applications? I mean, for me 581.29 driver is most stable, in games, but breaks youtube, any video player, MPV, etc!
Thats why I asked for different GPU people test my issue!

That's why I said "most driver issues are game specific". But I am using 581.42 with no issues so not sure what is happening there unless .29 has specific issues. I'm watching full 4k (3840x2160) 3D movies and running a remote desktop at the same time with my GPU regularly dipping well below 100% and no stutters so far.

No, 581.42 is the same as 29, different framepacing in different applications, which was the same for all applications for legacy 565.90 driver, 2 drivers before NVIDIA App transformation!

UPD: I will try dev drivers, 590.26 for test!

dawkinscm wrote:
RickyAstle98 wrote:
dawkinscm wrote:

If you are having issues with 566 and but are OK with 565 then that might be a local issue. 566.36 is probably the most well known stutter free driver across the Internet. I'm using a 570 driver and it's as least as good as 566.36 and I'm probably going to try later drivers now that I've sorted the stutter causing issues. I had local wifi and bluetooth issues. But I also recently found out that Windows 11 works differently to Windows 10 and "optimizing" it can make things worse.  I caused  much of the stuttering by turning off certain Windows CPU security features and hidden Nvidia features the GPU needs to work properly with Windows 11. Turning them back on helped to get rid of the stutters.

I need someone to test this with different or same GPU - my friends with almost identical PCs have same issues with different drivers, thats not local issue!

UPD: Thats VRR problem, simple FIX is running LSFG ontop of RIFE interpolation with same FPS gives very smooth behavior, which means the driver interferes VRR framepacing!

Told you lol. Most of the driver issues are either game specific or as you have found out, they are local setup issues.  BTW I'm now using the latest 58x driver with no issues. So it looks like in my case it was mostly local setup issues.

But how different drivers breaks framepacing in different applications? I mean, for me 581.29 driver is most stable, in games, but breaks youtube, any video player, MPV, etc!
Thats why I asked for different GPU people test my issue!

dawkinscm wrote:
RickyAstle98 wrote:

Mega Question about NVIDIA drivers / media player VSync behaviors

Stutters/jerks through any players after 581.X driver updates

Who can CONFIRM that any driver after 565.90 provides same/better smoothness levels?

As 4070 user, tested a lot different drivers, only 565.90 works flawlessly with any models, except 4.10ths and 4.25+

MPV behavior: 565.90 driver works flawless and perfectly smooth at any multiplier, w/wo interpolation, 566.36+ any drivers provides jerks!

UPD: Always running 565.90 driver before September driver release, changing TRT libraries do nothing! Any NVIDIA tricks like Fast Sync and others doesnt work, looks like new driver just breaks framepacing windowed modes at all, exclusive fullscreen works bad!

If you are having issues with 566 and but are OK with 565 then that might be a local issue. 566.36 is probably the most well known stutter free driver across the Internet. I'm using a 570 driver and it's as least as good as 566.36 and I'm probably going to try later drivers now that I've sorted the stutter causing issues. I had local wifi and bluetooth issues. But I also recently found out that Windows 11 works differently to Windows 10 and "optimizing" it can make things worse.  I caused  much of the stuttering by turning off certain Windows CPU security features and hidden Nvidia features the GPU needs to work properly with Windows 11. Turning them back on helped to get rid of the stutters.

I need someone to test this with different or same GPU - my friends with almost identical PCs have same issues with different drivers, thats not local issue!

UPD: Thats VRR problem, simple FIX is running LSFG ontop of RIFE interpolation with same FPS gives very smooth behavior, which means the driver interferes VRR framepacing!

Mega Question about NVIDIA drivers / media player VSync behaviors

Stutters/jerks through any players after 581.X driver updates

Who can CONFIRM that any driver after 565.90 provides same/better smoothness levels?

As 4070 user, tested a lot different drivers, only 565.90 works flawlessly with any models, except 4.10ths and 4.25+

MPV behavior: 565.90 driver works flawless and perfectly smooth at any multiplier, w/wo interpolation, 566.36+ any drivers provides jerks!

UPD: Always running 565.90 driver before September driver release, changing TRT libraries do nothing! Any NVIDIA tricks like Fast Sync and others doesnt work, looks like new driver just breaks framepacing windowed modes at all, exclusive fullscreen works bad!

Sach wrote:

Hello everyone,

I'm using Rife v25v2 TensorRt with scene change SVP Motion vector

I'm looking at the advanced settings of the software, all settings, it's been 3 months that I test several settings in a manual way directly, the options in the basic UI are really not sufficient, with the image comparison method I find that the image jerks despite having tested 50 different values.



So I wanted to know about the SVP motion vector method  in the all parameters tab, what are the best values to set to avoid artifacts between scene changes?



It's a pity there's no Discord because chatting on the forum is really hell for this kind of question.



and please don't tell me that the values in the drop-down menu directly in the profile are enough, because by exploring several softwares that don't work in real time there are so many options that you can modify with Rife and also use more efficient scene change methods, or use CVPR2023- A Dynamic Multi-Scale Voxel Flow Network for Video Prediction, or GIMM/GMFSS or MultiPassDedup

The developers of SVP4 arent responsible for the RIFE conf/params at all, you supposed to edit the contents of onnx models by yourself! Video jerks? Your issues, RIFE issues, not SVP issues!

Multiple Pass Deduplication is best scenario for any type of content, but requires large buffer to work realtime, also this is helps with old movies, when slowmotion scenes has duplicated frames or something else!

SeppoIisakki wrote:
RickyAstle98 wrote:

What means Lossless Scaling does not work at all in windowed mode?

Im Windows 10 and LS works perfectly well, even with RIFE combination!

I use RIFE x3 + LSFG (refresh rate) target, looks awesome!

The SVP doesnt support multiple instances, atleast RIFE workloads cannot be divided!

I tried the newest 3.something version lossless scaling - https://store.steampowered.com/app/9930 … s_Scaling/

When I enabled the scaling, it centered the video on screen and surrounded it with black borders. It did not stay windowed. I couldn't figure out what to do with it.

Good to know that multiple RIFE enhanced videos cannot be played.

In regular mode inceasing multiple video's fps seems to work: https://www.svp-team.com/forum/misc.php … mp;preview

I needed to adjust the "multiple" parameter, when I changed it from 2 to 4 in the SVP options, it allowed me to play multiple videos at 120fps. For example four videos at 120FPS use ~125W with a 5090.

Last LS talking Chainik, thanks!

The LS supposed to have Scaling type, and yes, you supposed to make windowed fullscreen video, better quality, doesnt work when you forced exclusive fullscreen mode through player!
Scaling type: Any (except OFF)
Scaling mode: Auto/Fullscreen

Chainik wrote:

SeppoIisakki

> I needed to adjust the "multiple" parameter

menu -> app. settings -> add. options -> running multiple videos

> Good to know that multiple RIFE enhanced videos cannot be played.

why? works for me...


RickyAstle98

> The SVP doesnt support multiple instances, atleast RIFE workloads cannot be divided!

not true for both parts...

BTW I'm going to start banning all the LS talks here big_smile

Maybe this is not true for 1st part, but my SVP4 refuses to work when RIFE workloads multipled, I dont know why!

SeppoIisakki wrote:

Does the SVP app support multiple videos at same time?

In other words, can I get it to simultaneously increase multiple playing video's FPS to ie. 120fps?

But when I open a new MPC-HC player the video smoothing "jumps" to the newly opened player, and previously opened video's FPS drops down back to default.

I'm on Windows 11. Otherwise it seems to work well, better than Lossless Scaling on Steam, which does not work at all in windowed mode.

E: Got the RIFE AI working too, uses around 200W of GPU power. Regular mode uses 60-70W without RIFE.

What means Lossless Scaling does not work at all in windowed mode?

Im Windows 10 and LS works perfectly well, even with RIFE combination!

I use RIFE x3 + LSFG (refresh rate) target, looks awesome!

The SVP doesnt support multiple instances, atleast RIFE workloads cannot be divided!

AutumQueen92 wrote:

On a 24fps video, there is not difference between 'to screen' and 'x6 framerate' right? Since my monitor is 144hz.
I can reliably hit that framerate when my 1080p videos are downscaled to 720p first via SVP

deleted

narkohol wrote:
RickyAstle98 wrote:
reconhell wrote:

Everytime i try to use rife it shows a CMD prompt. and it loads for a very long itme but my video never plays. This hasn't changed in 2 years. not with my rtx 4090 and rtx 5090. i dont get where you guys use this for... seems totally useless im sorry.

Try to change builder_optimization_level (SVP4/rife/vsmlrt.py) (default value: 3)


RTX 4070 timing cache
Level 1 - about 30s (attachment #1)
Level 2 - about 40s (attachment #2)
Level 3 - about 1m10s (default)

Line 181
Line 1863
Line 2021

I guess these modifications are not free and make the performance a bit worse or less 'optimized'?

±1.5%

reconhell wrote:

Everytime i try to use rife it shows a CMD prompt. and it loads for a very long itme but my video never plays. This hasn't changed in 2 years. not with my rtx 4090 and rtx 5090. i dont get where you guys use this for... seems totally useless im sorry.

Try to change builder_optimization_level (SVP4/rife/vsmlrt.py) (default value: 3)


RTX 4070 timing cache
Level 1 - about 30s (attachment #1)
Level 2 - about 40s (attachment #2)
Level 3 - about 1m10s (default)

Line 181
Line 1863
Line 2021

AutumQueen92 wrote:
RickyAstle98 wrote:

I recomend to use 4.9v2 legacy modes, always good for any type of movies, except VERY fast paced action, then 4.10+ and lite models!

The V2 models usually 15% faster than V1 models!

Also 4.26 about 1.3x more demanding than 4.9v2 (4070)

Thanks for replying. Are all the release files already v2? I don't see any file named 4.9v2

https://github.com/AmusementClub/vs-mlr … e_v2_v4.7z

All other releases have different folders, rife and rife_v2 for example!

As 4070 user, I always recomend V2 models, SIGNIFICANT increase in transcoding speeds - realtime improvement!

AutumQueen92 wrote:

Hello, I've recently gotten a 4060ti 8gb to replace my 2060s and RIFE seems to play anime quite well at 1080p x2=48fps. It looked amazing, but it really puts a heavy load on the GPU.

Are there any settings or files that anyone has tried that reduces workload?
I used the rife_v4.26 from releases in https://github.com/AmusementClub/vs-mlr … nal-models but nothing seemed to improve.

Settings are:
NVIDIA TensorRT
2 GPU threads
AI model: 4.26
Performance boost: Disabled and Enabled didn't change anything
Scene change: SVP motion vectors
Processing of scene changes: Repeat frame
Duplicate frames removal: Do not remove

I recomend to use 4.9v2 legacy modes, always good for any type of movies, except VERY fast paced action, then 4.10+ and lite models!

The V2 models usually 15% faster than V1 models!

Also 4.26 about 1.3x more demanding than 4.9v2 (4070)

Drakko01 wrote:
RickyAstle98 wrote:

Well, you prefer less compatible video player, my 180Hz GSync monitor works great, no matter what multipliers, targets being set!
I dont recomend RIFE x2.5 (24>60) then LSFG x2 because you will see scene transitions more often, since one half the frame being reprojected twice!

About TV VRR window, man thats just TV without LFC/SFD algorithms, which means LSFG will capture frames with JIT delays!
Anyway thanks for your testing, prefer RIFE x3 too, dont see any obvious for GPU cook, since LSFG paired w/RIFE frames allows super high refresh rate conversions with tiny amount of difference!

About tiny amount of difference, of course LSFG is bad from 24FPS conversion, but from RIFEd 48/72 - works/looks 4 times better!

I understand what you saying and I don't refute that a 180hz Gsync monitor can do much better than my tv at any multipliers. But I don't have monitors in my setup.

About Rife x2.5 I concur . whit the settings you suggested i try Rife x2+Lsfg 120hz and Rife x3 Lsfg 120hz

And for a more compatible video player. MPV for me is exactly the opposite, very jerky and stuttering playback .Maybe I change something and made mpv works like that. I really don't know.

If you have a way and time to tested mpv on a similar tv. Your recommendations will be appreciated.

About RIFE x2.5 - with float factors, LSFG generates 0.5th frame twice, thats why you will see scene transitions more often, thats not recomendation!

I dont have any jerk/stutter playback with MPV player, maybe you break something, perhaps! I dont have compatible TV for test, sorry!

Drakko01 wrote:

Thanks RickyAstle98/flowreen91  for your recommendations. Its none of that, the app capture fine the frames , enable Draw FPS its the first I do at the start of testing and its ok.

Close overlays like Steam, Discord, NVIDIA,  I never test something with overlays open!

I don't use Mpv, don't liked

The source problem, as I mentioned its hardware related, I think the tv accept better via Gsync compatible and VRR the 72fps or 96 fps of rife x3 and x4, the settings of the picture giving by RickyAstle98 give similar result but still  I prefer  Rife x3 /x4

I'm using windows 11 24h2 , mpcvr testing version, not the vapoursynth of svp, many variables to tweak and try

But again thanks both, I really enjoy testing others option to improve my viewing experience, and some many time through time rewatching movies or clip  I remember how we started and how far we've come today.

Many time thanks to this thread

Well, you prefer less compatible video player, my 180Hz GSync monitor works great, no matter what multipliers, targets being set!
I dont recomend RIFE x2.5 (24>60) then LSFG x2 because you will see scene transitions more often, since one half the frame being reprojected twice!

About TV VRR window, man thats just TV without LFC/SFD algorithms, which means LSFG will capture frames with JIT delays!
Anyway thanks for your testing, prefer RIFE x3 too, dont see any obvious for GPU cook, since LSFG paired w/RIFE frames allows super high refresh rate conversions with tiny amount of difference!

Abount tiny amount of difference, ofcourse LSFG is bad from 24FPS conversion, but from RIFEd 48/72 - works/looks 4 times better!

narkohol wrote:
Drakko01 wrote:
RickyAstle98 wrote:

[I tested RIFE x2 + LSFG 180FPS target (LSFG was updated and can generate frames without exact multipliers with consistent framepacing)!

LSFG from RIFE frames 48>180 looks almost the same as RIFE 24x7 and uses 6 times less resources!

Yeah I know, but i dont wanna pay for lsfg

LosslessScaling is the best $10 spent ever.

LS is the best $3 spent ever!