madVR - high quality video renderer (GPU assisted)

[Sunspark]

I like mitchell-netravali ar which uses the formula b=1/3 c=1/3 for chroma. It doesn't require much in resources and looks pleasant, good choice for integrated gpus. Not too saturated, a pleasant tonality. bicubic 60 ar (b=0, c=0.6) is similar, but a smidge more saturated.

MN is also good for image scaling too. I would argue for low resource systems that it's probably the most performant choice that actually does look good. Yes, for image scaling others like spline36 and lanczos3 are sharper, but you must ask yourself, should the image be that sharp? Sometimes too sharp can look worse.

If you have gpu to spare, then yes, use super-xbr as a better performer, but it will use more resources.

https://legacy.imagemagick.org/Usage/img_diagrams/cubic_survey.gif

[fabuloufab]

Hi Everybody,
(Very) long time user of MadVR here, I always found the information I needed about it on tutorials or reading other user’s posts in forum but this time I’m facing two strange issues that I can’t solve by myself and I need your ideas.

First, my setup :
- Hardware : Nvidia RTX 3060 Ti – Projector JVC NZ7 in 4K (Set in 120Hz)
- Software : Win 11 - MadVR 0.92.17 hdrMeasure205 – Latest Nvidia drivers (but result the same with plenty of precedent versions) – Lav filters 0.78 with Video hardware decode setting set to “none” (as latest 4k encodes regularly bug with hardware decode set to “copy-back”)

-First problem : - “The Image Enhancements” in the “Processing” section don’t apply on 4k sources.
I can enable all the filters and put them to the max, no filtering is done and I can’t understand why
As there is no scaling to be done (4k projected in 4k) I can’t use the “upscaling refinement” in the “Scaling Algorithms” but I would like to sharpen the image a bit

-Second Problem: -Either my projector or the Nvidia driver or MadVR bugs 50% of the time when switching back from BT 2020 to BT 709 after closure of a HDR movie resulting in a total black screen.
I want to use MadVR HDR tone mapping so on the “Calibration” section of my projector I did put “This display is already calibrated” with “BT. 2020” and I did tick “Report BT.2020 to display (Nvidia only)”
When launching the film, everything is great. The projector switch to BT 2020 and the image is wonderful. But when I close the movie, it only swith back correctly 50 % of the time. The rest of the time it seems to try to switch and my projector screen stay completely black until I either unplug/replug the HDMI cable or power off/on my projector (or pc).

Any idea would be welcome… ��

[Sunspark]

Regarding the 4K encodes bugging, you don't need to use copy-back on a system like that.. select the LAV decoder dropdown as D3D11, select the hardware device dropdown as "Automatic", set MadVR to use D3D11 for presentation and that should work better for you, no?

[Siso]

Regarding the 4K encodes bugging, you don't need to use copy-back on a system like that.. select the LAV decoder dropdown as D3D11, select the hardware device dropdown as "Automatic", set MadVR to use D3D11 for presentation and that should work better for you, no?

It will work better at the cost of presentation glitches.

[huhn]

Regarding the 4K encodes bugging, you don't need to use copy-back on a system like that.. select the LAV decoder dropdown as D3D11, select the hardware device dropdown as "Automatic", set MadVR to use D3D11 for presentation and that should work better for you, no?

this breaks deint and what has the presentation to do with decoding. madVR is dx9 with and without d3d11 decode or presentation

[Sunspark]

It would save the trouble of transferring 4K back and forth across the memory bus and instead work on the textures directly in the GPU.. at least this is my understanding of it.

The other stuff like presentation glitches or deint are separate issues..

For the glitches, maybe give it a try unchecking frames in advance mode so it uses backbuffers instead.. the old overlay mode used backbuffers.. maybe it would interact with the driver differently if it wasn't frames in advance. Who knows?

[huhn]

what has this todo with presentation.

if you get presentation glitches you get them not from decoding you get them from well presentation.

but d3d11 presentation has nothing todo with d3d11 decode.

[Siso]

what has this todo with presentation.

if you get presentation glitches you get them not from decoding you get them from well presentation.

but d3d11 presentation has nothing todo with d3d11 decode.

OK, this is starting to get confusing, what is the better way to set it up in lav video and madvr?

[huhn]

the way it works best for you. or sadly i can not give you a proper answer.

if copyback is not a performance issue for you use it.

if d3d11 gives you presentation glitches try d3d9.
if d3d9 still gives presentation issues try the presentation modes this mode has differences compared to d3d11 (it's been a long time since i tested it.)

[Siso]

No matter how many frames in advance I set up, it always gets presentation glitch or two. This is with d3d11 in lav, and d3d11 in madvr.

[fabuloufab]

Thanks for your answers,
As Siso said, Any hardware decoding done with Nvidia cards seems to give presentation gliches from time to time with latest x265 encodes so this is not an option.
Also I can't explain it, and it's subtle, but the visual quality of CPU (Non GPU hardware) decoding seems slightly better to my eyes.
But the problem is not there... It's with the two ones I mentionned earlyer.
Does somebody else use the "image enhancements" to filter source image that has the same resolution as your output and does it work ?
Does somebody else use the "Report BT.2020 to display (Nvidia only)” and experienced this problem ?

Thank you

[Siso]

Does somebody else use the "image enhancements" to filter source image that has the same resolution as your output and does it work?

Regarding this one, with my old monitor which had 2560x1080p resolution image enhancement didn't work on academy flat (1.85:1) sources, I don't know why...maybe madvr bug. On scope (2.35:1) and panavision (2.39:1) they did work.

[fabuloufab]

Ok, I just found out two interesting things for my 2 problems.

-First : I was wrong. The "Image enhancement is active. The thing is that while they worked great when I had a 1080p projector and wanted to sharpen and densify my 1080p sources, the filters doesn't seem tailored and adapted to 4k.
sharpen edges: Even at max level (4.0) No change is visible
crispen edges: Even at max level (4.0) No change is visible
thin edges: Change is visible but filter is not usable at all. As Asmodian says in his description: "As its name implies. Good for SD Anime or cartoons."
enhance detail: At max level (4.0) this one has a small noticeable effect that look nice... I would be interested to see if it could go higher... but it can't...
LumaSharpen: Does have an effect. Sharpen but quickly bring up "noise" to the image
AdaptiveSharpen: Does have an effect. Sharpen but quickly bring up "noise" to the image
activate anti-bloating filter: When you activat LumaSharpen or Adaptive Sharpen, this one is "proposed" automatically and annihilate everything done by these two above filters (That's why I didn't saw any effect)
activate anti-ringing filter: No change visible.

So now my question... Does somebody have something else to propose to filter a 4k image on a 4k projector that works better and is better adapted ?

-Second problem: I found out that the problem only occur if I diminish the movie image (Window mode) before closing mpchc-be with the x in the top right corner.
If I close the video right away from full screen with Alt-x or with the fullscreen flybar, It works. Strange...

[huhn]

don't combined them like that.
enhance detail is just a macro block enhancer if you run that one followed by adaptive sharpening it will just create block all over the image.

enhance details at 4 alone massively throws noise all over the image here at my side tested with a 4k source.

[Amuat]

Does anyone see noticeable benefits from reduce ringing artifacts? I've noticed that when I would really need it, it does absolutely nothing. On the other hand, it's good that I realized I was carrying dead weight on all my profiles.

[Heaud]

I have seen Reduce Ringing Artifacts work on sources that have haloing from being over-sharpened. The setting works, but it can also mistake some hard edges as halos and potentially reduce the contrast intensity of said edges.

[ryrynz]

Does anyone see noticeable benefits from reduce ringing artifacts? I've noticed that when I would really need it, it does absolutely nothing. On the other hand, it's good that I realized I was carrying dead weight on all my profiles.It's amazing on content that actually needs it. Always evaluate the options effects before selecting them.

[Amuat]

It's amazing on content that actually needs it. Always evaluate the options effects before selecting them.

Could you give me any examples when it has worked? Could there be some settings that prevent it from working in my case?

[ryrynz]

It's most noticeable on lower resolution content that's been upscaled/sharpened. For me taking this DVD quality content and upscaling it always made whatever haloing that was there so much more apparent. Having this enabled competely cleaned up the final image, giving me an upscale that looked naturally shaper without looking like it had gone through postprocessing.

[mzso]

Hi!

Was there some sort of genius timeout in madVR again? This time it seems to break the whole OS's ability to play videos, and not just crashes the mediaplayer. (The DVB app only shows it cannot connect the decoder to the renderer.)

[huhn]

hmm like a virus:
https://www.videohelp.com/software/madVR/old-versions

you most likely have 204 which timed completely today or something.

it's been like this for years now i guess you have to get used to it.
the older version crashed when the warning comes and you changed files this time you have the warning for a month and after that no more madVR.

some hotkey don't work with 205 have a pleasant day.

[mzso]

hmm like a virus:
https://www.videohelp.com/software/madVR/old-versions

you most likely have 204 which timed completely today or something.

it's been like this for years now i guess you have to get used to it.
the older version crashed when the warning comes and you changed files this time you have the warning for a month and after that no more madVR.

some hotkey don't work with 205 have a pleasant day.
Well, the last package I had saved was 169. But it didn't give warnings. I guess it could have been missing.

[mclingo]

Hi, I cant seem to get 24p to switch when i'm using this set of values in display modes below:

2160p23, 2160p24, 2160p25, 2160p30, 2160p50, 2160p59, 2160p60

If I delete these and just have 2160p24 it works

What am I doing wrong?

[huhn]

not playing a 24p files but a 23p?

[mclingo]

yes it will play a 23, 976 no problem, I even checked my TV would go in 24P mode using both Devcon and also display settings, I can't work out why it won't switch to 24P unless I only put 24P in display modes alone and delete everything else.

[huhn]

ok let's try again are you playing a 24p files a true rare one a 24.000 not 23.976. playing a 23.976 will not switch to 24p unless it has no other choice because that's wrong.

[movieman]

yes it will play a 23, 976 no problem, I even checked my TV would go in 24P mode using both Devcon and also display settings, I can't work out why it won't switch to 24P unless I only put 24P in display modes alone and delete everything else.

Check the frame rate of the file with Mediainfo.

https://mediaarea.net/en/MediaInfo/Download

[mclingo]

Yes I am playing a 24 Hertz file I don't have many of these

[huhn]

screen of the osd?

[huhn]

ctrl+j.
media info doesn't matter only what madVR get's as info matters and it will be in there.

[mclingo]

I think it was something to do with the way the particular set of videos I was playing were encoded as I found some other 24P videos that are older and they played fine so the issue is not MADVR, It was of the Mafia did recognise the videos as 24P as per osd but was playing them at 60p, I've got rid of them now cheers everybody.

[Sunspark]

So today I am playing with de-interlacing a DVD for the first time in a long time. I was trying different players, various hardware and software algorithms, etc. when I stumbled upon one that was noticeably clearer than the rest, though all did de-interlace.

If you have an old TV show (this one here was filmed in the 1970s) that was converted to DVD (which is interlaced), chances are high that it is using a 3:2 pulldown to make it 30i.

MadVR's deinterlacing option "force film mode" will use inverse telecine. "force film mode: force IVTC, reconstructing the original progressive frames from video encoded as interlaced, decimating duplicate frames if necessary. madVR's IVTC is not functional when using DXVA2 or D3D11 native GPU decoding because madVR's IVTC algorithm runs on the CPU and the decoded video data is never copied to system memory."

Video mode doesn't use IVTC.

In automatic it didn't seem to recognize the 3:2 and used video mode, but visually and logically I felt it must be using 3:2, so forcing it to film mode fixed that.

I also played around with adaptive QS HW deinterlacing and one caveat I'd make for that, other than that it's a little blurrier than MadVR's film mode, I believe it also manipulates the chroma because even though it still says SMPTE C in the HUD and while the output is fine, it was slightly hotter than MadVR's software-only mode. It's something to be mindful of because it means chroma will get touched twice unless you use dxva chroma or whatever.

[huhn]

video mode in madVR is pretty much the same as QS deint and the deint in other player because it is just DXVA 2 deint.

these usually also IVTC but they don't decimate.

[Sunspark]

In today's installment of 'twiddle settings':

These old TV series that were filmed on 24 fps film and converted to 29.97 NTSC DVDs via 3:2 pulldown, for best playing results when using madvr's "film mode" deinterlacing for inverse telecine (video mode doesn't have ivtc), change the Hz of your display to 48 Hz even though the material is reported as 29.97 instead of leaving the display at 60 Hz. Internally the original frames are 24 fps, and if your display is at 48, it'll sync up better.

The internal rate of 24 being produced can be confirmed by the fact that smooth motion will turn itself on at 50 and 60 Hz, but off at 48 Hz which proves that there are 24 frames not 30.

Smooth motion at 60 makes it a little too smooth due to the nature of it blending adjacent frames together. This is better than interpolating a brand new frame like TVs do (awful) but it's still a blending. I usually have this feature off, but it was helpful for testing purposes here.

I think for 720x480 DVDs, it is arguably better to upscale using madvr to native resolution and 48 Hz, than fighting with custom parameters to output 720x480 24 Hz. You can do this, but the downside is that it will need to be upscaled anyway and it's a question mark whether the panel can do a good job. On my old monitor, it does not do a good job upscaling from 480 to 1152. The second challenge is that square aspect pixel ratios on old stuff, madvr needs to downscale the image in the X dimension to make the presention right. It is blurrier than if madvr does the scaling to native itself because you would end up scaling twice. Once downscaling in X to fix the aspect ratio, and then a second time in X and Y to fit the panel by the panel. Better to just do one scaling pass instead of two using the algorithm of your choice. Making the panel work at 24 is iffy (custom parameters again).

[huhn]

every working hardware deinterlacer is ivtcing this is not a joke.
it's even part of the DXVA2 spec a driver can even tell if it is able to do that or not. the problem is they don't decimate they simple can't anyway. the supported IVTC modes can literally be listed in the DXVA info.

madVR will automatically use 23p in film mode even with 2:2 material if possible or 120 48 what ever.

smooth motion is incapable of making something smoother then the original frame rate in the best of all cases it looks the same. if it is smoother your display is broken like frame skipping or broken 23p a Samsung special.

720x480 and 720x576 and it doesn't matter because these are not display resolution by itself because it can be 16x9 or 4x3 you can technically output without scaling.

madVR never does simple scaler in one go and the assuming that doing something one time is better then otherwise is false. madVR does x first followed by Y which will literally result in the same result as a one pass scaler. why someone would ever let a panel scale a custom resolution beyond me anyway it will be bilinear and no surpirse for everyone.

[Klaus1189]

Samsung special - 23p broken? What product was affected?

[huhn]

it's an old story back in the native 60 HZ panel time and back when they said they have like 9000 hz samsung was often incapable of doing 23p natively they just did 3:2 even on some close to flagship TVs. while LG and sony where able do that even on 60 Hz panel (maybe they just run at 48 i don't know) as long as the TV was not an entry class TV they could do it.
that was the same time where panasonic was able to do 23p correctly but only outside of PC mode it was glorious adding 3:2 (6:4 in this case) to a native 120 HZ panel marble of engineering.

an example is the KU7000.

BTW. samsung LG and sony can turn 60p with 23p in it into 23p in real time. times change.

[Sunspark]

Complicated, and stuff is different from setup to setup. Great example, I have a Steam Deck and last night I wanted to see what AMD's deinterlace looked like (in Linux). It looked horrible, very low-resolution like an old video game. Completely unwatchable. I looked it up and saw some thread complaints here and there talking about how AMD is in the process of losing hardware support for de-interlacing (including in Windows), which means on AMD any support for it moving forward is going to have to be a software-only solution, I guess to go along with the pre-existing hardware decoding chroma issue.

My panel is Samsung manufactured and is from 2010, so this might be the same time period huhn is referring to.. I'll have to try again to get that madvr test motion clip to work, I tried in the past but it didn't play at all.

In Windows, my igpu is Intel (and really, for htpc this is probably one of the better solutions since it has working chroma and deinterlacing). Flossycake gave me a telecine test clip and the results were interesting.. there is a test pattern at the beginning then people in the next section (I think the cadences are changing too). I played it through a bunch of stuff to see the results as I noticed the section with people after the test pattern had combing in it.

madvr film mode (runs on cpu), combing
madvr video mode (runs on gpu), no combing
lav's adaptive hw qs w/ madvr deint off, combing.
lav's software algos w/ madvr deint off, combing.
mpc-be's decoder w/ evr-cp in dxva-hd mode, no combing.
mpc-be's decoder w/ mpc-vr in software mode, combing.
mpc-be's decoder w/ mpc-vr in hardware dxva mode, no combing.
mpv default settings, combing.
vlc auto, no combing.
microsoft films&tv, no combing.

I was surprised that madvr's film mode failed and produced combing because it has decimation and the other stuff.. but there was a real difference with this test clip.

The original reason I got into letting the panel scale was because of less system resources needed since I would only scale chroma and mostly because it supported 23p, 24p, 25p, etc. in 1920x1080 out of the box without my having to make custom modelines to make it work at native resolution. When I use 1920x1080 at 23p or whatever, the MadVR HUD tells me frame drop is on the order of days. Is the HUD accurate? I don't know. It claims everything is fine, but it doesn't mean it is if the panel is lying. I don't even need to scale, I could also just set the aspect ratio to 1:1 in the menu instead of fill or aspect which just means a smaller picture without any image scaling performed. But is it truly 23.976? the MadVR HUD claims it is, but I just don't know. Accepting a signal doesn't mean it is processing it as expected. I am able to overclock (I don't) this panel past 60, but when I do, it still displays a picture but it starts dropping frames visibly.. motion becomes choppy.

I'm still suspicious about smooth motion. It is frame blending, the image is being manipulated. I have read that one might see blur on edges in motion which you wouldn't see on a "true" vsync. Both setting 48 w/o smooth motion and and 60 w/ smooth motion look similar, but I still feel the smooth motion looks a tiny bit too smooth (though not as bad as the soap opera effect).. in other words, my eyes feels the movement of the people isn't quite fast enough or separated enough. Feels like I have to work harder watching.

So yeah, I'll have to try again to see if I can make that test clip work. Too many variables in play.

[Sunspark]

Finally, I have solved it.. the reason madTestPatternSource (link is in the first post of this thread) doesn't work is because it's a 32-bit app.

In order to make the .ytp files work without replacing your existing 64-bit install, what you need to do (beyond the initial install of the testpatterns) is you need to grab the 32-bit version of mpc-hc or whatever 32-bit version of the directshow player you use, not the install .exe but the .zip instead, e.g. "MPC-HC.2.1.7.2.x86.zip". Then you can unzip to a folder, run the exe from the folder and drag the .ytp files to it and it will play without replacing your existing 64-bit install.

[Sunspark]

Ok I need help with the test pattern.. what am I supposed to be seeing?

Using "SmoothMotion24.ytp" as an example seems to look the same whether I have smooth motion on or off at native resolution/edid refresh. Changing it to 1920x1080@24 and running it in a window so no image scaling is performed still looks the same whether I have smooth motion on or off as it does in native.

Am I looking at the wrong thing? What am I supposed to be noticing? The top bars look blurred no matter what.

Update: Ok, I can see it now. I think my panel is a fake 23,24,25, etc. With the test pattern, if I block the bottom 3 bars with my arms and look at the top 3 moving at say, 48 with SM off, I can see judder (or stutter?).. at 59.876 with SM on, it doesn't have the same kind of judder, BUT it looks like text is vibrating on a plate.. this is more pronounced if you again use your arms and look at the 3rd row up from the bottom.. more blurry at 60 w/ SM on than at 48 w/ SM off. So, the tradeoff is sharper text clarity with some slight judder vs less judder (or more constant steady pulsed judder) with blurrier text as a result.

[Sunspark]

Happy Saturday, a special treat today for tweakers.

A 4K 16-bit PNG greyscale gradient that was programmatically generated WITHOUT dithering, displaying ranges from 1-bit to 12-bit.

https://imgur.com/fxILke1

[anta777]

What is the most profitable video card to buy now for high-quality upscaling?
Which card has the best ratio fps/price?
AMD or Nvidia?

[huhn]

these are used cards and been used means the prices are all over the place.

even a what is the fastest let's say 200 euro GPU is hard to answer but still easier then this.

Happy Saturday, a special treat today for tweakers.

A 4K 16-bit PNG greyscale gradient that was programmatically generated WITHOUT dithering, displaying ranges from 1-bit to 12-bit.

https://imgur.com/fxILke1
ok 12 bit with massive banding i'm supposed to look at what? if my rendering surface is 8 bit the result will be 8 bit or 12 on an 10 bit, 8 bit or 6 bit display because the display doesn't matter even if the image viewer is not dithering the GPU driver will and even if the GPU driver doesn't have to dither the display will. making this totally pointless. getting 12 bit out of a computer start to end is already close to pointless.

here you go have a 16 bit gradient: https://www.bealecorner.org/red/test-patterns/Gradient-16bit.png
now you just take the combination that gives you the least banding does this tell you the bit deep of something no because nothing can for 100%.

[Sunspark]

@huhn The purpose of the file I linked is that it helps you figure out what the bit depth of your panel is, or the display chain. The file you linked doesn't do the same thing, it won't tell someone if they have 6 bit or 8 bit or 10 bit panel, just that they have banding or no banding. It's useful to know what the panel presents as if you turn off dithering in the GPU and player (e.g. 10 bit).

[huhn]

as i said it doesn't because a picture can't do that.

if a image viewer rounds to 8 bit you will get 8 bit on any display even 6 bit panels.
this can at beast show you the bit deep of an image renderer that's it.

the file i linked allows you to test bit deep in details. because the display bit deep doesn't matter anyway only the end result.

you can easily test 6 bit GPU dithering with that file doing 8 or 10 bit on the software side and other such things because it is a perfect banding free 16 bit gradient.

every panel accepts 8 bit input even when it is 6 bit all of them dither this to 6 so it does nothing. even if you manage to find a panel that is not dithering it has to be like 20 years old or broken.

[Amuat]

What is the most profitable video card to buy now for high-quality upscaling?
Which card has the best ratio fps/price?
AMD or Nvidia?

I'd like to expand and modify this. What are the most important stats on a GPU (VRAM, memory bus, etc.) in terms of madVR performance? Am I looking at different aspects compared to gaming performance, or is it pretty much the same? And what is the relative role of CPU, or RAM for that matter?

[huhn]

gaming is an indicator for the same type of GPU like all 4000 series nvidia card can be roughly compared like this.

make sure it is not ray tracing DLSS and such.

Vram barely matters. 4 G should be able to do most stuff 8 Gb should be able to do anything.
ram also doesn't matter you just need enough so the system works 16 will do fine for HTPC video only.
CPU also don't really matter make sure it has 16 lanes and a modern PCIe spec like 4

[Drybonz]

I have some videos that have some deinterlacing line artifacts that the deinterlace options in MadVR, in any configuration, can't seem to eliminate. Are there any any additional options to watch these with more clarity? Thanks for the help.

[Amuat]

I got a new PC with the aforementioned GPU (previously 1070), and I can't use display modes in madVR anymore. It just gives me black screen, no matter if HDR is on or off, or the video is SDR or HDR. In fact when I tried it with HDR off (on SDR video), I didn't just get black screen, but broke everything. I wasn't able to use HDMI optimal mode anymore or the monitor disconnected, only standard mode. In the end the only thing that fixed it was reinstalling Windows from scratch. Now everything works normally again, without the display modes of course, have to manually change to a matching refresh rate every time I watch a video, as the stuttering is horrendous even with VSync applied in the control panel (smooth motion certainly changes things, but not all for the better, really weird feel to it). I also tried to change the color from 10 bit to 12, and my monitor went absolutely berserk. It didn't even respond to remote control commands at first, just a circle flashing like crazy at the corner. I'm lucky I was able after multiple restarting of the PC to change it back when the image sort of came back in this sort of drunken double vision thing with constant flashing.

What is going on here? My monitor certainly supports HDMI 2.1 and I never had any problems with it on my old PC, the cable is one of the best in the market and short enough, and I obviously have Windows updated and all the latest drivers. Games work perfectly on full screen and with the now added VRR and G-Sync support. And here I was thinking how nice it would be to finally have the HDMI support my monitor was made for, and also a little more powerful GPU on top of that. Is there something inherently problematic with the new 40 series, do I have to change some settings in Windows/Nvidia Control Panel/MPC-HC/madVR, or what? The whole thing seems extremely fragile now, and I feel like I'm walking constantly on egg shells in regards to video rendering. Some help would be extremely appreciated.

[Zoltrix]

Hi all,

Been using MadVR for a while with no issues. One of the recent software updates introduced occasional black flickering. See video clip below (around the 7-6 sec mark as the video counts down) for an example. Nothing at all changed on my setup, just an update to MadVR. Since then, I rebuilt the HTPC (fresh OS install), tried a different video card, using Potplayer instead of MPC-BE. Issue still exists.

I can go into more detail. But based on the video clip shown, is there a setting or something to eliminate this flicker?

https://imgur.com/a/R05GPPr