madVR - high quality video renderer (GPU assisted)

[ryrynz]

For this decoding method, fan speeds were the same and system load was lower for D3D9 compared to D3D11.

By "system load" you're referring to what exactly?

[Sunspark]

By "system load" you're referring to what exactly?

The burden on the CPU and GPU, the higher the load the more heat/rpm and slower responsiveness. The lower the % the better.

But you know what, I was just watching something, and I feel madvr even in dxva scaling mode provides for a fractionally sharper image/better colour than the corresponding equivalent in mpcvr. The difference is subtle, but I feel it is there.

[ryrynz]

The burden on the CPU and GPU, the higher the load the more heat/rpm and slower responsiveness. The lower the % the better.



You would need to measure actual wattage of the system. Anyway, most will choose what works for them and their systems, no real need for comparisons.

[flossy_cake]

37% combined load is quite low and if your fan is spinning up & making objectionable noise then the cooler is woefully inadequate. System responsiveness should not be affected, something isn't right. Check clock speeds -- maybe it's throttling? Is it a laptop?

[huhn]

the problem is percent of what?
CPU and GPU are boosting and lowering the speed as they see fit so a percent number doesn't mean much anymore.

[Sunspark]

It's basically a laptop, this is a tiny little NUC. It's a small cube-shaped PC, sort of like a Raspberry Pi, but more powerful. I don't mind if the fan is running, but there is a certain threshold where if it goes past a certain point you hear "Eeeeeeeeeeeeeeeeee". The fan is a little blower fan. I've thought about replacing it, but the replacement would still be a blower of the same size. People getting the steamdeck are also complaining about fan noise too from 1 of the 2 fan suppliers. I am considering the steamdeck as my next PC.

CPU: Zen 2 4c/8t, 2.4-3.5GHz (up to 448 GFlops FP32)
GPU: 8 RDNA 2 CUs, 1.0-1.6GHz (up to 1.6 TFlops FP32)
APU power: 4-15W

[flossy_cake]

the problem is percent of what?
CPU and GPU are boosting and lowering the speed as they see fit so a percent number doesn't mean much anymore.

If it's downclocking then real usage % is even lower and still shouldn't be spinning the fan and creating system responsiveness issues. But I always disable downclocking on the CPU & GPU as it can cause dropped frames. AMD is particularly bad in this regard as their driver doesn't necessarily offer a reliable way to disable it and I had to use a janky app called ClockBlocker which works by creating a small, persistent artificial workload on the GPU to trick it into not downclocking.

[flossy_cake]

Just some comments about MadVR's sharpening filters.

I find the "adaptive sharpen" algorithm seems to do a good job of sharpening content which was produced in SD, but was upscaled by the mastering studio to HD, where the studio did not add any sharpening of their own. This kind of content seems difficult to sharpen as post-resize sharpening tends to need higher strength and seems more prone to bloating/ringing artefacts as a result. This is where adaptive sharpen seems to shine -- it seems to detect where the true edges are, even though the source is pre-upscaled and all the edges have soft interpolated pixels around them.

Otherwise if the source is typical native SD or HD, I find the "luma sharpen" algorithm nicest & is my "go to" sharpener if I feel the content looks soft. Alternatively the "sharpen edges, crispen edges, enhance detail" seems to do a good job too, although with ringing. The "thin edges" option I avoid using as I've seen it dramatically altering the overall shape of objects and text.

The debloating option appears to negate the sharpness filters too much so I don't use it. The deringing option seems to barely affect perceived sharpness so it's usable.

I don't consider myself a sharpness junkie, but I am definitely not one of these videophiles who sets their TV sharpness to 0 unless I'm at very close viewing distances eg. PC monitor type distance. I am currently using a 51" TV @ 3.5m viewing distance. This will probably be replaced with a 55" at some point and then I'll probably have to go back and readjust all my sharpness profile strength values. Visual acuity will obviously play a big part as well.

[scollaco]

Just some comments about MadVR's sharpening filters.

I find the "adaptive sharpen" algorithm seems to do a good job of sharpening content which was produced in SD, but was upscaled by the mastering studio to HD, where the studio did not add any sharpening of their own. This kind of content seems difficult to sharpen as post-resize sharpening tends to need higher strength and seems more prone to bloating/ringing artefacts as a result. This is where adaptive sharpen seems to shine -- it seems to detect where the true edges are, even though the source is pre-upscaled and all the edges have soft interpolated pixels around them.

Otherwise if the source is typical native SD or HD, I find the "luma sharpen" algorithm nicest & is my "go to" sharpener if I feel the content looks soft. Alternatively the "sharpen edges, crispen edges, enhance detail" seems to do a good job too. But the "thin edges" option I avoid using as I've seen it dramatically altering the overall shape of objects and text.

The debloating option appears to negate the sharpness filters too much so I don't use it. The deringing option seems to barely affect perceived sharpness so it's usable.

I don't consider myself a sharpness junkie, but I am definitely not one of these videophiles who sets their TV sharpness to 0 unless I'm at very close viewing distances eg. PC monitor type distance. I am currently using a 51" TV @ 3.5m viewing distance. This will probably be replaced with a 55" at some point and then I'll probably have to go back and readjust all my sharpness profile strength values. Visual acuity will obviously play a big part as well.


Are you applying these as post image enhancements or part of the upscale refinement for e.g when upscaling 1080p to 4K?
For me, upscaling 1080p to 4K I use sharpening filters as part of the upscale refinements and then for native 4K on a 4K display, I might add a little bit of post image enhancement like Sharpen Edges and enhance detial.

[flossy_cake]

Are you applying these as post image enhancements or part of the upscale refinement for e.g when upscaling 1080p to 4K?
For me, upscaling 1080p to 4K I use sharpening filters as part of the upscale refinements and then for native 4K on a 4K display, I might add a little bit of post image enhancement like Sharpen Edges and enhance detial.

I am just using them under image enhancements & I'm not sure what the difference is vs upscaling refinement & haven't done any screenshot comparisons to tell.

I'm actually scaling everything to 1366x768 as that is the resolution of my TV :o

MadVR seems to have its own internal logic regarding whether it applies the sharpening pre/post resize depending on how close the source res is to the final output res, so if you're tuning sharpening settings I would recommend doing it in full screen mode (or whatever you use most of the time).

[x7007]

Everything works the first time and than Bit Depth stop working, Black Level and many others, changing them doesn't change anything on the image. I can't understand these in windows 11. using the madvr 159 version, it happens on every one of them. Deleting madvr registry settings makes them work again.

[Sunspark]

Is the behaviour different on Windows 10?

[kasper93]

here have fun
if AR was available it was used:
NN: https://abload.de/img/nntiksf.png
bilinear: https://abload.de/img/bilinearmaj1e.png
bicubic: https://abload.de/img/bicubicijj8s.png
biliteral sharp: https://abload.de/img/biliteralsharpfrj7e.png
biliteral soft: https://abload.de/img/biliteralsoftuijwi.png
jinc: https://abload.de/img/jinct0kri.png
lanczos 3: https://abload.de/img/lanczos9eknq.png
NGU AA: https://abload.de/img/nguaaq6kw4.png
NGU sharp: https://abload.de/img/ngusharp1ij39.png
reconstruct sharp: https://abload.de/img/reconstructsharpemkd8.png
reconstruct soft: https://abload.de/img/reconstructsofth6kvo.png
super xbr 100: https://abload.de/img/superxbrr5k0i.png
superres 4bilinear: https://abload.de/img/superresbilinear4tqjic.png

my comments: super xbr looks like it is just the same as NGU AA in this test while been far cheaper.
NGU sharp looks quite aliased and also fuzzily the lines are not strength and when compared to reconstruct sharp it shows how to get the image that sharp without that issue. the source is quite aliased for chroma so that maybe the reason why both are quite aliased but "correct" in this case.
superres bilinear 4 is failing in this test it adds an dot over the first a there is some hints of chroma there but only it amp it to eleven.

reconstruct sharp is the winner for me in this test. if you can't stand the aliasing the biliteral results are also very promising on this single test.

Thanks, that's nice comparison. I was wondering if there is existing work with more examples. This example shows nicely how good is biliteral sharp, when there is aligned luma channel to make it work. And it works better the better luma is.

Generally biliteral is my favorite, because when it works it works nicely, and when it fails it is in scenarios where there is no sharpness in the first place, so it is not that noticeable. For sure NGU or even bicubic is safe consistent choice, but for sharp content with text or some overlay I like the biliteral feel.

For softer "movie" content chroma upscalling is generally not that noticeable, unless it is not broken...

[el Filou]

The ideal way to test chroma upscaling would be with a sample you know for sure to be from a native 4:4:4 capture, and there's not a lot of them around.
It's also important to use a real scene if you mostly watch live action content. If you only watch anime then anime is good but something that gives good results on it won't necessarily give realistic results with other content.
When I did my comparisons a few years back I used 4:4:4 and 4:2:0 versions of a sample showing an aerial view of a seaside town (can't remember where it came from, EBU or ITU maybe?), with various shades of red/orange/yellow on the buildings, and Reconstruction Soft gave the most fidelity (i.e. closest to the native 4:4:4) for me, but unfortunately the sample was only 720p so I'm not sure it would be the same with 2160p.

[Balling]

Prores software decoding can still be software directly rendered. I just brought it as example, because that uses the best Intel's Y416. There may be bugs in madvr, pixel formats in ffmpeg are bitperfect.

[James_b]

Hi Guys! how do you use sharpening? sharpen edges is extremely heavy compaired to LumaSharpen. Espesially if you running interpolation in the mix.

[Sunspark]

The ideal way to test chroma upscaling would be with a sample you know for sure to be from a native 4:4:4 capture, and there's not a lot of them around.
It's also important to use a real scene if you mostly watch live action content. If you only watch anime then anime is good but something that gives good results on it won't necessarily give realistic results with other content.
When I did my comparisons a few years back I used 4:4:4 and 4:2:0 versions of a sample showing an aerial view of a seaside town (can't remember where it came from, EBU or ITU maybe?), with various shades of red/orange/yellow on the buildings, and Reconstruction Soft gave the most fidelity (i.e. closest to the native 4:4:4) for me, but unfortunately the sample was only 720p so I'm not sure it would be the same with 2160p.

Logically, it would be the same since resolution is just a raster grid. If you established between two identical samples of the same resolution differing only in chroma that reconstruction soft gave you the most fidelity on your hardware, then the same should hold true at 2160p between two samples of the same resolution as well.

[flossy_cake]

Hi Guys! how do you use sharpening? sharpen edges is extremely heavy compaired to LumaSharpen. Espesially if you running interpolation in the mix.

I made some sharpness presets for each algorithm. I tuned them using SD sources and some soft looking HD sources, as those are the only sources I wanted to add sharpening to.

I tried to make it so the presets have similar strength across algorithms. eg. my LumaSharpen medium strength preset would have similar looking strength as AdaptiveSharpen medium strength.

I chose actors' faces as the main test pattern as that's what I'm looking at most of the time. If I had chosen some bright highlight detail like fence rails or text I may have ended up with different values.

LumaSharpen
low: 0.65
medium: 1.30
high: 1.95
max: 2.60

Comments: LumaSharpen seems to be a good all-rounder for lots of content, but seems less effective at post-resize sharpening. MadVR should be applying it pre-resize (at least when in fullscreen mode) so this shouldn't be an issue unless you are trying to use it on content which has resizing or blurry upscaling baked into the source.

AdaptiveSharpen
low: 0.2
medium: 0.4
high: 0.6
max: 0.8

Comments: seems to be better than LumaSharpen in that it is also effective at post-resize sharpening, turning very soft edges into sharp ones. With very soft pre-upscaled anime like Samurai Champloo Bluray it can turn the outlines into sharp HD-like lines at high strength values, without ringing artefacts.

CustomSharpen (sharpen edges, crispen edges, enhance detail)
low: 0.5, 0.5, 0.5
medium: 1.0, 1.0, 1.0
high: 1.5, 1.5, 1.0
max: 2.0, 2.0, 1.0
all customsharpen presets: anti-bloating 100% strength & anti-ringing ticked

Comments: seems to target high contrast edges only, making them appear thinner, sharper and more round, giving the appearance of a higher resolution source. Seems to need a sharp source to work well & doesn't seem to have much effect if used post-resize or if there is blurriness baked into the source.

To use these presets on a per-file/folder basis, create the above profiles in the image enhancements section, then tag a parent folder or file name with eg. [profile='AdaptiveSharpen medium'].

Alternatively you could write a script under MadVR's "profile autoselect rules" (see madvr's tips.txt file for details). But I didn't find this useful as it can't differentiate between files which have the same resolution but different sharpness baked into the source. I find the tag is also a handy reminder every time I open a file to let me know a sharpness profile is going to be applied.

I'm still deciding whether I like LumaSharpen or AdaptiveSharpen more. I haven't seen any source which didn't work well with AdaptiveSharpen so maybe it is the better one.

I haven't played with "SuperRes" as that is only part of the upscaling settings which would only get applied when upscaling.

[Siso]

Did you apply any AR and AB settings?

[flossy_cake]

Did you apply any AR and AB settings?

Only as specified in my custom sharpen presets. If I didn't specify a setting it means I didn't enable it.
Anti-bloating seems to almost completely suppress LumaSharpen and AdaptiveSharpen, but not the CustomSharpness ones (sharpen edges, crispen edges, enhance detail).

edit: but for "image upscaling" settings I'm using Lanczos 3-tap with anti-ringing ticked.

[James_b]

Thank's Flossy!
Any idea how to make 2:35:0 / 1080p for ex look good on a filled 16:10? Got no experience using "upscale refinement" tab.
The best i got without peak(GTX960) seems to be croma upscale: Cubic. image upscale: Lanczos.

[flossy_cake]

Thank's Flossy!
Any idea how to make 2:35:0 / 1080p for ex look good on a filled 16:10? Got no experience using "upscale refinement" tab.
The best i got without peak(GTX960) seems to be croma upscale: Cubic. image upscale: Lanczos.

It depends how you want the image to look... I'm guessing you're zooming 2.35:1 to 16:10 and not satisfied with the sharpness of the 800p resolution?

Just tried it now with a couple 2.35:1 movies and it really seems to depend how much sharpness is baked into the movie itself. One movie looked fine zoomed in but another looked a bit soft. Just pause on a scene and toggle it on/off in MadVR settings to compare (takes effect when hitting apply).

Personally I seem to like the medium presets. I've updated the AdaptiveSharpen presets after doing some more comparisons, lowering them a little to match the LumaSharpen ones. I think they should be staggered by either 0.2 or 0.25, but it's a tough call as they work in different ways on different scenes compared to LumaSharpen.

[x7007]

Is the behaviour different on Windows 10?

I dunno if it's different, having only Windows 11. not sure if it's AMD drivers or Windows 11 or Madvr. it's a mess.

[Balling]

Here is 4:4:4 10 bit from RGB 64 bit (16 bit per component, 4 components, RGBA). https://drive.google.com/drive/u/0/mobile/folders/113tgFZecNF1rFcSW5Pv6VjbSK1hAXwJH

Requires Y410, at least 30 bit per pixel.

mpv.com -v --hwdec=nvdec -gpu-api=vulkan --vo=gpu "01. test444_23.976.mp4"

Will cause HW decode on 20 series Nvidia and later. No HW decode with --hwdec=d3d11, alas. IT IS POSSIBLE with those patches https://patchwork.ffmpeg.org/project/ffmpeg/list/?series=527 at least on Intel® UHD Graphics 770 (12th gen BigCore).

Madvr does not HW decode.
Davinci is HW decoded too, since it does not use d3d11, it uses directly nvdec.

[ashlar42]

here have fun
if AR was available it was used:
[snipped long list of pictures]

my comments: super xbr looks like it is just the same as NGU AA in this test while been far cheaper.
NGU sharp looks quite aliased and also fuzzily the lines are not strength and when compared to reconstruct sharp it shows how to get the image that sharp without that issue. the source is quite aliased for chroma so that maybe the reason why both are quite aliased but "correct" in this case.
superres bilinear 4 is failing in this test it adds an dot over the first a there is some hints of chroma there but only it amp it to eleven.

reconstruct sharp is the winner for me in this test. if you can't stand the aliasing the biliteral results are also very promising on this single test.
Thanks for this. I've been using super xbr for a long time, best quality/performance ration in my opinion.

[James_b]

Is "reconstruct sharp" a new feature? I'm using 127.0.0.1 atm. Is it heavy like the super xbr?

[flossy_cake]

Is "reconstruct sharp" a new feature? I'm using 127.0.0.1 atm. Is it heavy like the super xbr?

127 is an IP address, version is at bottom left corner. Should be 0.92.17 if you're on the latest version, which I am using, and yes reconstruct chroma seems heavy: +20% GPU load on GTX1070 @ max clock. Super-xbr does not increase my GPU load at all compared to Cubic and Bilateral.

[tp4tissue]

Recently been using just the vcgt, and honestly, sometimesss...... not all times, just some times, I really enjoy the super exaggerated colors on wgamut. For example gilligan island and star trek tos, their shirts are basically 240 r, 0 g, 0 b, this looks crazy red without 3dlut.

with 3dlut, that pulls inward with less purity.

[x7007]

If I enabled the Windows 11 OS HDR I always get 10bit .. anyone have an answer for the 10bit issue stuck to 8bit after quitting a game or something and restarting Explorer.exe?

[tp4tissue]

After testing many consumer oriented monitors/ tvs, NONE have very good 10bit performance.

Only prosumer ipses have remotely good 10bit, and ips = bad contrast so they're just not good for movies.

Stick with 8bit and use madvr's dithering. Unless you're not tonemapping, in which case, it doesn't matter.

Going in and out of 10bit is always going to be buggy, because the interplay causes the GPU and TV to do some weird stuff sometimes.

[tp4tissue]

The ideal way to test chroma upscaling would be with a sample you know for sure to be from a native 4:4:4 capture, and there's not a lot of them around.


The ideal way to test chroma is DON'T BOTHER.

Waste of time, waste of Gpu power.

Impossible to notice from real viewing distance on real content, you can only see it on very specific test patterns.

[tp4tissue]

IMAGE QUALITY IS DAMAGED IN P010 MODE!



So, dxva pipeline optimizations can't support every mode encoders will use.

They can only account for the most common formats. You shouldn't be watching trashy 10bit anime encodes anyway. 8bit jpn bd_remux or it's not even worth watching. :D

If there's banding in the source, I want to see it. those 10bit debanded vaseline encodes = image corruption.

One could legitimately say, the use of p010 is the problem, not the other way around.

[huhn]

should i mention UHD BD are 10 bit and p010 is absolutely fine?
or maybe i hope he just get bored and leaves?

[flossy_cake]

Supposedly 10bit encodes better than 8bit, even though the source is only 8bit. I don't know if it's necessarily true, but some of my 2mbps 10bit x265 1080p files look surprisingly decent for such little data. Grain is suppressed though, and for some reference movies I'm not satisfied.

Bigger problem seems to be storage — if I want a large collection at high bitrate / remux, I cannot find a HDD large enough and quiet enough to hold it all.

[huhn]

The ideal way to test chroma upscaling would be with a sample you know for sure to be from a native 4:4:4 capture, and there's not a lot of them around.
It's also important to use a real scene if you mostly watch live action content. If you only watch anime then anime is good but something that gives good results on it won't necessarily give realistic results with other content.
When I did my comparisons a few years back I used 4:4:4 and 4:2:0 versions of a sample showing an aerial view of a seaside town (can't remember where it came from, EBU or ITU maybe?), with various shades of red/orange/yellow on the buildings, and Reconstruction Soft gave the most fidelity (i.e. closest to the native 4:4:4) for me, but unfortunately the sample was only 720p so I'm not sure it would be the same with 2160p.

you want a 4.4:4 source?
here we go:
native: https://abload.de/img/hatrednpjc2.png
lossy 4:2:0 encode: https://drive.google.com/file/d/13wodrGah8JVBVVM6APmXNjLksWhLu-gI/view?usp=sharing
bicubic: https://abload.de/img/bicubicl5jkc.png
biliteral sharp: https://abload.de/img/biliteralsharp64jyq.png
biliteral soft: https://abload.de/img/biliteralsoftbmkyt.png
NGU AA: https://abload.de/img/nguaa5okra.png
NGU sharp: https://abload.de/img/ngusharpufjur.png
reconstruct sharp: https://abload.de/img/reconstructsharp4gk4q.png
reconstruct soft: https://abload.de/img/reconstructsoft9bk40.png
super XBR: https://abload.de/img/superxbrlgjz4.png

take the HUD not the game content didn't work out as planned.


i wonder why the colors are messed up lavfilter was used to subsample BTW..
Thanks for this. I've been using super xbr for a long time, best quality/performance ration in my opinion.
well said.

Thanks, that's nice comparison. I was wondering if there is existing work with more examples. This example shows nicely how good is biliteral sharp, when there is aligned luma channel to make it work. And it works better the better luma is.

Generally biliteral is my favorite, because when it works it works nicely, and when it fails it is in scenarios where there is no sharpness in the first place, so it is not that noticeable. For sure NGU or even bicubic is safe consistent choice, but for sharp content with text or some overlay I like the biliteral feel.

For softer "movie" content chroma upscalling is generally not that noticeable, unless it is not broken...

there are other chroma tests good luck finding them here.
it has been done a lot.

[tp4tissue]

Supposedly 10bit encodes better than 8bit, even though the source is only 8bit. I don't know if it's necessarily true, but some of my 2mbps 10bit x265 1080p files look surprisingly decent for such little data. Grain is suppressed though, and for some reference movies I'm not satisfied.

Bigger problem seems to be storage — if I want a large collection at high bitrate / remux, I cannot find a HDD large enough and quiet enough to hold it all.

the debanding algo works better with the 10bit pipe. so the encoder is presuming that that's what the audience wants smooth gradients. If that's true, go for it. but purists go for bd_remux

The hdd space is a matter of income inequality, eat the rich.

as for huhn being a di(k as per usual, I'll leave it at that as I can't understand what he says half the time, so I forgive him based on the language barrier, as he probably doesn't fully grasp what we say either..

[huhn]

i'm hurt by your fantasy facts:
https://x264.nl/x264/10bit_01-ateme_pierre_larbier_422_10-bit.pdf
https://x264.nl/x264/10bit_02-ateme-why_does_10bit_save_bandwidth.pdf
https://x264.nl/x264/10bit_03-422_10_bit_pristine_video_quality.pdf

but keep making stuff up and keep calling me a dick because you like to make stuff up.

[mclingo]

I find anything but a 4k bluray remux looks garbage, not everywhere but in enough places.

[flossy_cake]

the encoder is presuming that that's what the audience wants smooth gradients. If that's true, go for it. but purists go for bd_remux

I'm pretty sensitive to video filters so I reckon I'd know if some debanding filter was applied. So far I can't detect anything, but I can detect the film grain is compromised and I think that is more a function of x265 encoder settings and/or the 2mbps bitrate, rather then some pre-filter by the human who created the file.

But I have noticed some people are baking into the file their own degrain, sharpen, and even S-curves, and I find that really upsetting. I had to scrap around 200GB of files as they were contaminated with such filters. Even more upsetting is the comments section where users are praising the uploader saying how much it "pops".


The hdd space is a matter of income inequality


If money was no object I would still have a problem as the WD Red Plus 12TB 3.5" drives appear to be the largest size with acceptable noise levels to me, and I would still need to buy at least 5 to house an entire collection of remux, and the noise would be multiplied by 5, and I probably don't have enough SATA ports. I think the solution would be to have a separate room with a rack full of HDD's in RAID and access them by ethernet in the HTPC room.



eat the rich


https://i.kym-cdn.com/photos/images/original/001/252/520/f83.jpg

[tp4tissue]

I find anything but a 4k bluray remux looks garbage, not everywhere but in enough places.

2000% agree..

I love anime, but 1080p, crappy adobe aftereffects stitching. blechhh...

[tp4tissue]

i'm hurt by your fantasy facts:
https://x264.nl/x264/10bit_01-ateme_pierre_larbier_422_10-bit.pdf
https://x264.nl/x264/10bit_02-ateme-why_does_10bit_save_bandwidth.pdf
https://x264.nl/x264/10bit_03-422_10_bit_pristine_video_quality.pdf

but keep making stuff up and keep calling me a dick because you like to make stuff up.

Huhn literally misreads 50-60% of what people posts and responds with something he invented in his own mind. I guess no one's told him. But seriously huhn, you either have a psychological disorder, or some sort of trauma that's painted the whole world an enemy.

Half the time, I see you arguing with people that isn't even talking about the things you're talking about, JUST like how you're arguing with me about something completely irrelevant to the point i was making.

This is a seperate issue from you being a legit di(k though, I want to believe it's a language barrier, but again and again, you purposely draw negative attention on yourself.

Are you a bad person? Maybe not... But you need to go back to some sort of finishing school to learn manners.

[tp4tissue]

I'm pretty sensitive to video filters so I reckon I'd know if some debanding filter was applied. So far I can't detect anything, but I can detect the film grain is compromised and I think that is more a function of x265 encoder settings and/or the 2mbps bitrate, rather then some pre-filter by the human who created the file.


Anime encoders are extremely liberal with debanding filters. It's used excessively.

Anime is typically extremely low budget, so their post is very quick in very basic softwares, there's gradient ramps Everywhere. Ever since 10bit became popular with encoders, these ramps have all but disappeared, but if you download some 8-bit remuxes, you'll see them. there're still in the vast majority of modern animes.

[huhn]

very convincing.

so you didn't wrote this none sense?
the debanding algo works better with the 10bit pipe. so the encoder is presuming that that's what the audience wants smooth gradients. If that's true, go for it. but purists go for bd_remux
just attack me personally more and avoid answering.

10 bit encodes are not better because deband works better in pipeline and because the encoder "presuming" something.
the encoder can just do better work that's why 10 bit is "better".
the rest is what you made up in your mind.

BTW. what has 10 bit encoding you bring up here todo with madVR at all.

nice high ground you take by attacking me personally and about bad manner BTW.

maybe go for another a spyder x is useless or just keep about offtopic "anime" encodes.

another run of me been a dick and bad manners and i'm not even sorry.

[tp4tissue]

very convincing.

so you didn't wrote this none sense?

another run of me been a dick and bad manners and i'm not even sorry.

Huhn, you're arguing with yourself, I don't even know what you're mad at. You quote people and you've misread their posts under a completely incorrect context. and Yes, half the time you are a di(k, like right now.

I've made none of those points that you think I made. You've read those quotes completely wrong.

In YOUR post, you're talking about encoder as in the video encoder.

In MY post, I'm talking about the encoder as in the guy who's encoding and deciding to use deband too liberally destroying certain details.

And NO, this is not an isolate incident or an honest mistake, YOU ACTIVELY look for these fights, picking apart statements completely inappropriately to draw attention onto yourself.

[huhn]

this is what you quoted:
https://forum.doom9.org/showpost.php?p=1968293&postcount=63039

yes i'm reading a question if x264 10 bit is better than 8 bit wrong.

have a good read here:
https://forum.doom9.org/forum-rules.htm

edit:
In YOUR post, you're talking about encoder as in the video encoder.

In MY post, I'm talking about the encoder as in the guy who's encoding and deciding to use deband too liberally destroying certain details.

And NO, this is not an isolate incident or an honest mistake, YOU ACTIVELY look for these fights, picking apart statements completely inappropriately to draw attention onto yourself.
and again personal...

[tp4tissue]

Huhn, you've read it wrong. I don't know how many other ways there is to say it.

And don't throw the rule book at me, you break the rule all the time with your negative holier than thou attitude mouthing off at people presuming to "know better".

The worst part of it being, again, you've read their post completely wrong, and you're not even in a real argument, yet you gaslight people all the time.

[huhn]

ok let me help you another way:
:readrule:

[tp4tissue]

Right back at'cha

:readrule:

[flossy_cake]

Anime encoders are extremely liberal with debanding filters. It's used excessively.

Maybe it works better with anime due to the lack of texture details compared to the real world? Are you able to show a comparison? I'm interested to know if I could spot it in a double-blind test, or whether I might subjectively think it's "better". I guess it would depend if texture detail is lost.

[James_b]

Hello pro guys! If i export a video in an editor with "Format profile: Main@L5" Will that produce a better image than "Format profile: Main@L4"?