[ / / / / / / / / / / / / / ] [ dir / agatha2 / b2 / choroy / dempart / thebestb / vg / vichan / webcams ]

/a/ - Animu & Mango

Winner of the 80rd Attention-Hungry Games
/otter/ - Otter For Your Soul

THE INFINITY CUP IS COMING BACK
May 2019 - 8chan Transparency Report
Name
Email
Subject
Comment *
File
Password (Randomized for file and post deletion; you may also set your own.)
* = required field[▶ Show post options & limits]
Confused? See the FAQ.
Embed
(replaces files and can be used instead)
Oekaki
Show oekaki applet
(replaces files and can be used instead)
Options

Allowed file types:jpg, jpeg, gif, png, webm, mp4, swf, pdf
Max filesize is 16 MB.
Max image dimensions are 15000 x 15000.
You may upload 5 per post.


Welcome to /a/, please read the rules before posting.
Reminder that in the event 8ch goes down, our bunker will still be up and running.

File: 7a30743f315a999⋯.png (275.11 KB, 512x512, 1:1, 7a30743f315a999d715a3b204b….png)

File: 684a7718aef7080⋯.jpg (3.15 KB, 200x200, 1:1, 684a7718aef7080ab1bd60ea54….jpg)

 No.928144

ITT: share software that are used to manage or share your stash of anime

I will go first, Hydrus is pretty good for downloading 8chan thread and image galleries. >>>/hydrus/

Also for those who watches cartoon mpv would be good

Question 1: Is Waifu2x cheating?

Question 2: Anyone here uses ffmpeg?

 No.928149

>Is Waifu2x cheating?

Considering how well it works I'm pretty sure whoever made it signed some demonic pact.


 No.928154

>>928149

> Demonic Pact

The Waifu2x devs admitted that they can make it even better by adding more neural network layers but "it might get too slow", so really it is just Rick & Morty esque "fuck god" technology (I regret saying this)


 No.928158

what mpv config files are we using these days?


 No.928161

File: 79df490259198de⋯.png (783.76 KB, 1280x720, 16:9, [HorribleSubs] Endro - 03 ….png)

>>928154

I regret you saying that too.


 No.928163

>>928161

I study machine learning as a CS student, and unfortunately the technology felt ungodly. Like... really ungodly.

>>928158

Also questtion: Are the the MPC-HC + MadVR fags still here?


 No.928165

File: f50b020aebc8eae⋯.jpg (443.2 KB, 1280x720, 16:9, [Erai-raws] Hitoribocchi n….jpg)

>>928154

>(I regret saying this)

This isn't a face to face conversation. Just hit the backspace button.


 No.928167

>>928165

It's a regret that has to be said, like a dead son or daughter.

There is a poetic element to this.


 No.928169

>>928144

For those who do not want MPV, and want the bonus of installing system-wide codecs, download K-Lite Mega Codec Pack.

The player is MPC-HC.


 No.928171

>>928163

Didn't development for MPC-HC end? What's the official weeaboo-seal-of-approval project now?


 No.928172

>>928163

>Also questtion: Are the the MPC-HC + MadVR fags still here?

You mean anyone 18+ who has been watching anime for more than a decade? No. These people left the board along with its quality.


 No.928173

>>928171

It didn't end. In fact, it ended and returned.

The project will always be K-Lite.


 No.928175

>>928169

>>928173

So new name?


 No.928176

>>928175

K-Lite Codec Pack is 15 years in constant development and updating.

Mega version is recommended.

https://codecguide.com/


 No.928178

>>928169

Or be cool like me and have both.

Just be careful not to accidentally install the adware shit that is bundled with K-lite installation. Read the menus carefully.


 No.928180

>>928171

From what I understand, development ended, but not updates or support. MPC-HC is still completely fine to use and probably will be for years.


 No.928186

So, Waifu2x just blows up your image? Why would anyone use that?


 No.928187

File: a2dae032af621f2⋯.png (217.89 KB, 634x910, 317:455, 021.png)

File: 126cfd8eec81973⋯.png (3.85 MB, 1268x1820, 317:455, 021.png)

>>928144

>Question 1: Is Waifu2x cheating?

No. It's amazing, and works really well for scanlations of low resolution web manga. While it does have a bit of an issue with the text, it still is preferable to low resolution manga.

>>928178

Nani? Are you sure your file is clean? Post the SHA256 or MD5 hash, some of the non official site mirrors might bundle their installation file with adware.


 No.928188

>>928186

Because you have a thing you want to clean and typeset but the only raws available are very low res.


 No.928190

File: fe15c27789296ae⋯.jpg (241.28 KB, 711x986, 711:986, 022.jpg)

File: 06d9b8e1f0655d9⋯.png (3.08 MB, 1422x1976, 711:988, 022_waifu2x_art_noise3_sca….png)

>>928187

Maybe post an example where it actually works well. That second image got smudged to hell. Try multiple noise reduction settings.


 No.928192

File: 08e6a5ce125783d⋯.png (4 MB, 1268x1820, 317:455, 021.png)

File: 73fb6f90f076982⋯.png (3.91 MB, 1268x1820, 317:455, 021.png)

File: 0c6fbffc20a27bb⋯.png (4.02 MB, 1268x1820, 317:455, 021.png)

>>928190

>Try multiple noise reduction settings.

Sorry, posted the wrong image. I still haven't cleaned up my Somali upscale folder, and out of habit posted the level 3 denoise version. Here's level 1,2 and no denoise.


 No.928198

File: c33e7f068fffdc4⋯.png (739.42 KB, 602x1003, 602:1003, Screenshot_1574.png)

File: c23b3e18ccf0217⋯.png (1.28 MB, 1773x1008, 197:112, Screenshot_1519.png)

>>928167

No, it really didn't. Rick and Morty overall had little to no correlation to what you were talking about that couldn't have been explained by different means, you just felt the need to insert it or even worse it's the first thing that came to your head. However, being self aware of where you are you included the regret as a way of self-defense so no one went too hard on you. If you're going to be annoying with your analogies at least do it properly


 No.928200

>>928178

>Just be careful not to accidentally install the adware shit that is bundled with K-lite installation. Read the menus carefully.

There's no such thing, I just installed it.


 No.928201

I double clicked the OP in the catalogue and all I got was this thread instead of the latest episode of My Hero Academia.


 No.928207

File: 1bb3fd5895f11e9⋯.jpg (36.63 KB, 340x351, 340:351, 1428300712885.jpg)

>>928163

>Are the the MPC-HC + MadVR fags still here?

I'm still using CCCP.


 No.928208

>>928207

That's what I used before I found out how much better mpv was than MPC-HC.


 No.928211

File: 4ba8abd3dbbb154⋯.jpg (115.33 KB, 1920x1080, 16:9, mpvsnap-ゴブリンスレイヤー #10 (00:….jpg)

>>928158

ewa_robidoux with mitchell params for downscaling (they're both the exact same filter with different params set, and using ewa_robidoux with mitchell params is how you get mitchell + EWA for higher image quality on non-integer downscales), though personally I just go ewa_robidouxsharp for downscaling because the sharpness looks nice.

ewa_lanczos or ewa_lanczossharp (one isn't clear cut better than the other both visually and on benchmarks) for upscaling.

If your hardware can't handle sinc scalers (such as lanczos), go ewa_robidouxsharp. If ewa_robidouxsharp is still too heavy, go mitchell. If mitchell is too heavy then anything but a simple bilinear filter is too heavy, go bilinear.

Leave deband on always, mpv's debanding is amazing. If you're really concerned about things being mis-debanded then use settings more conservative than the defaults, but the defaults are near perfect and guaranteed to do more good than harm.

Get some neural network upscalers/downscalers. You'll need a high end GPU. I use a gtx 780 and have performance issues if I use fbo-format=rgba32f, but fbo=format=rgba16hf can do neural network downscaling for even 5k+ video and upscale even dvd resolution (720x480) video with the shaders I suggest.

At the very least use KrigBilateral, it's a neural network chroma upscaler/downscaler and the best out, it's also very lightweight for upscaling (but not for downscaling).

There are better luma upscalers than FSRCNNX, but FSRCNNX is the best luma upscaler that at least a single GPU currently out can run. Go for FSRCNNX_x2_8_0_4-1 on most high end GPUs (this is what I use on my gtx 780), and FSRCNNX_x2_16_0_4-1 if you have the latest and greatest.

For luma downscaling use SSIMDownscaler always.

If you want to squeeze the absolute highest image quality for every kind of video, get the auto-profiles.lua script. You can tell it to use different settings on different resolution ranges.

Here's my mpv.conf file as reference:

https://bpaste.net/raw/db59bd93bb13

Also consider the set of scripts which turns mpv into an image viewer. In my experience that's actually the only good image viewer in the market, it's the only one with good format support, and it also has all of mpv's tricks for image quality:

https://github.com/occivink/mpv-image-viewer


 No.928215

>>928211

This /tech/nician gets it. Robidoux is incredible when you see the lack of halo it provides.

I'd add

[/code]

video-sync=display-resample

interpolation=yes

[code]

to get the equivalent of madvr's "smoothmotion".

Don't forget to get hyped for the new version that totally revamps vulkan support.


 No.928216

>>928215

Well, MPC-HC has it already. And LAV, and HW Acceleration, and shell integration, and an actual UI.


 No.928248

>>928198

> Rick and Morty overall had little to no correlation to what you were talking about

But it does. As I said in >>928163 I was trying to use that as a reference point on how unnatural yet predictable it is.


 No.928289

File: 875415a40ad9373⋯.jpg (32.26 KB, 218x281, 218:281, Hatsuni Miku eating a rand….jpg)

>mpv fags still screeching about their shitty player

>mfw installing LAV Filters Megamix and just using good ol' MPC-HC


 No.928291

I often watch movies on my home theater system in my TV by connecting my laptop, and when I tested MPV, it didn't had 7.1 or 5.1 passthrough. Worthless.

MPC-HC has the 'same as input' option and work wonders.

Also, _I can actually change the subtitle style_.

I don't know if anything has changed, but MPV is too green.


 No.928295

>>928291

>when I tested MPV, it didn't had 7.1 or 5.1 passthrough

But it does.


 No.928303

https://bigjpg.com/ is neat, in the vein of waifu2x


 No.928328

>lusers not knowing that LAV is just ffmpeg for windblows

Pure gold.


 No.928335

>>928163

> Are the the MPC-HC + MadVR fags still here?

MPC-BE + MadVR is better.


 No.928345

File: a714db17a39b3c0⋯.png (271.22 KB, 469x458, 469:458, i see a nigger.png)

File: 86a96ff08194ece⋯.png (2.53 MB, 1573x951, 1573:951, ui.png)

>>928172

>i just watched yurucamp in mpc-be+madvr

>after watching 2160 rips of netflix nature shows in hdr on my oled tv

<board quality is low

Leave.

>>928335

someone beat me to it

>>928171

This is the MPC-BE ui after a couple settings changes. What you can't see is my cursor hovering in the upper right corner to bring up the semi-transparent menu bar. Very clean interface,full-featured, and perfect for shut-ins who always have their blinds drawn.


 No.928348

File: 39db39ca39e01f5⋯.webm (15.97 MB, 3840x2160, 16:9, Obuchi Sasaba and Mt. Fuj….webm)

>>928345

>using netflix instead of watching 2160p rips of glorious nippon nature scenery directly on this board

SHAMEFUR DISPRAY


 No.928352

I only use HOW_DID_I_PLAYED_BACK.txt

That's MPC-HC + MadVR and I reinstalled everytime I had to reset my computer and it's workin perfectly fine.


 No.928366

File: d94dea49f2e0014⋯.jpg (1.51 MB, 1920x2160, 8:9, d94dea49f2e0014bde3f25f54c….jpg)


 No.928402

File: f66aaa4ebc78438⋯.png (1.4 MB, 1520x1080, 38:27, 1.png)

File: 61eb9e21f84f563⋯.png (1.69 MB, 1520x1080, 38:27, 2.png)

>>928345

I started using MPC-BE too. It's good, but I found an annoying problem with how the default subtitle renderer will always make the subs look pixelated and with rough edges in screenshots, but looked totally fine in playback. I managed to fix that by installing VSFilter. First screenshot is internal subtitle renderer, second is vsfilter. It's not as noticeable here, but it was worse in some other shows.

Can't see a point in MadVR. Looks the same, but my player is slower.

>>928366

Is that an upscale? Bottom one kinda looks like if you installed a bad Reshade set up with too much bloom, saturation, sharpness and FXAA on a PC game. The eyes of the black haired girl and the lips of the blue haired girl look weird in the MadVR example.


 No.928404

>>928402

You're retarded. There's no fudging in the image, the MPC-HC + madVR is the original look, and the VLC one is demonstrating its grey tinting and inability to reproduce 0-255 or anything 8bit and beyond.


 No.928413

>>928404

I didn't say VLC looked better, I'm just saying that MadVR doesn't look as right as it should either. Maybe it is the original look, I don't know nor care, but in that case it's just the source render that is shit.


 No.928414

>>928402

> Can't see a point in MadVR

It's an extremely powerful tool, but you have to configure it right. It wouldn't make you any good out of the box, though. The most useful feature is neural network upscaling, it makes 480p dvd-rips of older anime a lot more tolerable, and with some hard filtering you can make even shitty VHS-rips filled with noise look less bad. For the modern 1080p bd-rips it almost makes no sense, they are good enough as they are and can hardly be improved, unless it's a particularly shitty rip of an unpopular franchise.


 No.928427

File: 1a363cdd32232aa⋯.jpg (299.76 KB, 704x1616, 44:101, QUALITY_down.jpg)

>>928402

When doing any comparison it needs to be from the same frame. With mpv, if you have the OSD enabled, you can click the current time to display sub-seconds. VLC may or may not have something similar but you can frame advance and alt-tab to mpv to find the match anyway.

>>928414

Comparison screenshots from you too would be nice. While there's a strong difference in >>928366 also make them individual shots and ideally PNG so you can just flip between them to see, or even open them in an image editor with difference layering.

>>928348

Oh christ I would've assumed he meant the number of rips. But if you're using stream sources for anything above 720p, you're doing it very wrong. Even then, I've seen things from smaller sites where the video turns to scrambled dogshit if there's a lot going on. It's unacceptable. This is also ignoring the video degradation Netflix sometimes do if under load.


 No.928434

>>928427

> Comparison screenshots from you too would be nice

Comparison between what and what? MadVR and MPV? I made it a year or so ago, and MadVR was obviously superior. MPV-fags started screeching, that if you configure it right, it would be better, so I fear the same thing would happen this time too. You can share your configs with me, if you want a fare comparison.


 No.928435

>>928434

Ultimately it is irrelevant to me since my system doesn't usually handle 1080p even with mpv's XV output. But if you're saying x is better than y, comparison images need to be provided and easy to toggle just because of all the snake oil and outright garbage video filters that exist. That cuts both ways mind for configured mpv vs. MadVR.


 No.928438

>>928163

>Are the the MPC-HC + MadVR fags still here?

Yes, but I had to leave it because linux didn't support those. These days I use mpv and am fine with it.


 No.928439

>>928158

If you want your mpv screenshot filenames to be MPC-HC-like, stick

screenshot-template="%f_snapshot_%p_[%tY.%tm.%td_%tH.%tM.%tS]"
in your mpv.conf.


 No.928441

File: 5fcc02528df0814⋯.jpg (190.98 KB, 1080x1080, 1:1, sakura worry.jpg)

>>928435

>my system doesn't usually handle 1080p

Is everyone here on a toaster?

t. 50 fps 1080p AV1 CPU playback via dav1d


 No.928488

>>928435

>my system doesn't usually handle 1080p

Have you tried --gpu-dumb-mode=yes with the gpu VO?

Straight up general purpose computing inside the GPU is a fucking terrible meme every single GPU ever is terrible at, and this option forces a strictly fixed function pipeline with no compute shaders as god intended.

The downside is that most gpu vo options are silently disabled as most of the fancy stuff mpv does uses compute shaders, but the performance is massively improved.


 No.928490

>>928435

If you have an older PC, do not download anything HEVC/x265 or VP9. Prefer the eternal AVC/x264 and you'll run 1080p just fine if you enable hardware acceleration.


 No.928527

>>928435

Just buy yourself a newer PC. Mine is not that good, just 9700k and RTX 2080, but I never have problems with watching video, even with neural network supersampling.


 No.928528

>>928527

It's not the right time. Wait for Zen 2. I myself will wait for Zen 3 or 4.


 No.928589

>>928488

Doesn't seem to make a difference on or off, unless the option's been removed. Also I do have mpv and quite a few major dependencies built from source with -O3 -march=native and apparently I need to add libx265 to that since I have anime encoded with it.

>>928527

It's not a question of affordability. Anything past a certain point can't be trusted at a hardware level. Also how good is hardware support when it comes to anime on the PC side, because even modern smartphones don't bother supporting 10-bit colour hardware-wise, and most video players there don't even support anything fancier than yuv420p except for mpv. Even VLC is dodgy there.


 No.928590

>>928589

>most video players there don't even support anything fancier than yuv420p except for mpv. Even VLC is dodgy there.

MPC-HC supports RGB-32. YUV is fast but color accuracy suffers tremendously.


 No.928601

File: cd33a52ea7e9933⋯.png (1.89 MB, 1280x720, 16:9, Osomatsu-san.S01E01.720p.T….png)

It's just a shitty filter. It won't improve image quality, it just changes it to something else, which you may or may not end up preferring for periods of undisclosed time.

Anime doesn't have highly detailed backgrounds so the resolution you're watching something in barely ever matters. There's no trees swaying or strings dangling and barely is a director ever even bringing such shots in view to begin with.

tl;dr 480p is plenty, waiting for blu-rays is a waste and anything past episode 1 might as-well be presented in audiobook format.


 No.928603

File: 32722b8d84b3852⋯.png (1.05 MB, 1280x720, 16:9, [HorribleSubs] Zombieland ….png)

>>928601

>480p is plenty

Are you an ant?


 No.928608

This has gotten me curious. Is there a way to pre-render waifu2x and other upscaling algorithms if your PC can't handle it in real-time? Are there encode groups that bother with this?


 No.928609

File: e632a2d2f262ae0⋯.jpg (63.97 KB, 770x563, 770:563, 1428644198218.jpg)

>>928601

>tl;dr 480p is plenty, waiting for blu-rays is a waste and anything past episode 1 might as-well be presented in audiobook format.


 No.928621

File: 7da0f11af97316b⋯.png (1.21 MB, 847x1200, 847:1200, aqua.png)

>>928601

That is very studio-dependent. A few actually do good animation work, and master at decent resolutions too. Still I don't generally like to futz with video and there's enough re-masters out there that just make the colours look wrong and such.


 No.928642

>>928589

> Also how good is hardware support when it comes to anime on the PC side, because even modern smartphones don't bother supporting 10-bit colour hardware-wise

Everything starting nvidia Pascal cards (gtx 1060 and likes) support 10bit H.265 on hardware level. Older cards like GTX 960, don't support it. And it's a really bad idea to watch anime on smartphone. It's a complete disrespect for a media and creators.


 No.928664

>>928601

>>928603

720 > 480 and 1080


 No.928673

>>928664

You're quite right, anime is produced in 720p, so 480p is downscaled (which actually ends up looking better than it should) and 1080p is meme upscaled.

>>928642

>And it's a really bad idea to watch anime on smartphone. It's a complete disrespect for a media and creators.

Your opinion, my phone has a 6 inches screen and 2K resolution, the PPI is high and image quality is outstanding, better than the PC or TV. Also, the latest Qualcomm processors have hardware acceleration for 10bit HEVC.


 No.928675

>>928673

> anime is produced in 720p

How far do I have to go til anime is Standard Resolution?


 No.928676

>>928608

>Is there a way to pre-render waifu2x and other upscaling algorithms if your PC can't handle it in real-time?

Yes. Except you'll have to whip it up yourself with shell scripts.

You can also give mpv an absurd swapchain depth, It'll eat a lot of ram and be reset every time you seek.

>Are there encode groups that bother with this?

No.

>>928589

If you press shift+I while playing something with mpv you can at least tell if it's a decoder or an output bottleneck. Because with or without the GPU vo you have performance trouble, I think your processor just can't decode h264 in real time.

If it's decoder MPV has a bunch of options for speeding it up (at the expense of compatibility/image quality, also codec dependent):

--vd-lavc-skiploopfilter

--vd-lavc-fast

If you really want to squeeze the most performance out of everything install gentoo so you can compile it all from source, also try using link time optimization on mpv.

You can also compile a kernel with simple ticks (not dynticks, not tickless). This provides a large performance boost because it won't allow your CPU to idle, whereas on other modes linux fights the userspace on whether the CPU should be idling or not, which lowers performance.

And consider this kernel patch: https://github.com/graysky2/kernel_gcc_patch/ . You can pick march=native for the kernel through menuconfig with it.


 No.928677

>>928675

TV channels are either 720p or 1080i. Until TV gets a better transmission technology so that it can transmit at better resolutions, colors and even audio quality, anime will always be 720p.

I'm of the opinion that we should've get rid of TV and Radio frequencies and make the Internet a large uniband for Wi-fi, TV, Radio and Mobile Network. VoLTE is already a thing, goddamit, and so are internet radio and IPTV/streaming services.

Why this crap is still staled is beyond me.


 No.928678

>>928677

I am referring to "old anime" in Standard Definition before the 720p or 1080i shift


 No.928679

>>928673

No, phones are objectively worse. You need big screen to see more details. Plus, frame by frame rewind of great moments isn't even possible in the most of phone video players, and storage space is limited. The only thing phones are better is licking a photos of an anime girl you like, they need less cleaning after that.


 No.928680

>>928673

> 1080p is meme upscaled

You literally can't watch fullscreen 720p video on the screen with bigger resolution without upscaling the video first in real time. And upscaling it beforehand gives opportunity to use better upscaling algorithms, that are impossible in real time, and keeps you PC more cool, so not downloading the biggest resolution available is doing yourself a disservice.


 No.928681

>>928680

1080p takes up a lot more hard drive space. For a barely noticeable difference in upscaling quality, I don't think it's worth it.


 No.928682

>>928680

>>928681

How about pre-cache before playing, so when you are watching the video the system tries to catch up with the upscaling?


 No.928683

>>928681

So what? HDD are cheap, just buy yourself more. You are not a true otaku if you sacrifice a quality of anime viewing.


 No.928684

>>928683

I am also a CS student you know, I would rather use Machine Learning to make anime 1080p mp4/webm/mkv files a history


 No.928685

>>928684

They will never be obsolete. Machine learning never guarantees 100% accuracy.


 No.928686

>>928685

Bitch is that a challenge?


 No.928687

>>928686

It's the definition of Machine Learning. You should repeat a year, if you don't know even that.


 No.928688

>>928680

They just use your tired bilinear. 720p + madVr is a better solution.


 No.928690

>>928687

I know it can't be 100 but 99.999% would be enough for me to tell you to go fuck yourself.


 No.928691

File: 316b75afb7e0349⋯.png (2.22 MB, 1920x1080, 16:9, 720p.png)

File: f4276d3554ac672⋯.png (2.26 MB, 1920x1080, 16:9, 1080.png)

>>928688

No, 1080p + madVR is better. As You can see, lines are fucked up in 720p.


 No.928693

>>928690

99.999% isn't good enough. Even if the single pixel is not the same as creators wanted it to be, it's a betrayal. Tens of people worked hard to bring you an ultimate masterpiece and you changing it with some shitty algorithms. Do think yourself a better anime creator than them?


 No.928694

>>928693

Do we need a turing test for anime super-resolution?


 No.928699

>>928691

720p uses Nearest Neighbor + madVR

1080p uses a hard encoded Bilinear + madVR.

That's why it looks "better". On emulators they do as well. The different lies in the Nearest being sharper only.


 No.928707

File: 904c42fdd14720f⋯.png (1.04 MB, 1280x720, 16:9, [Erai-raws] RobiHachi - 03….png)

File: 055400795cf0354⋯.png (2.1 MB, 1920x1080, 16:9, [Erai-raws] RobiHachi - 03….png)

>>928677

>Why this crap is still staled is beyond me.

Because digital TV broadcast standards still use MPEG-2 for whatever reason.

>>928681

Why aren't crunchyniggers using NN scaling algorithms for their "1080p" encodes?

FSRCNNX werks quite well on current day 720p TV anime.


 No.928710

>>928699

They both use NGU-sharp very high, though.


 No.928716

File: bcdca00534431b8⋯.png (484.52 KB, 848x480, 53:30, 1.png)

>>928676

>Yes. Except you'll have to whip it up yourself with shell scripts.

How would one (me) go about this?

>>928681

Right. So why aren't you using 480ph265xaac? Fucking mentally ill hoarding hypocrites I swear.

>>928691

I can see but it's still the same anime. I don't see the big deal. You're still watching a trashy A-1 anime (I can tell by the bland background art). Why do you care about watching shitty mouthflaps at a higher resolution? The fuck is wrong with you people? Why are you watching shitty ads then take some moral highground of quality?

>>928707

Not a blu-ray. Why even care about quality if you're gonna half-ass it by watching a shitty tv broadcast? Terrible show by the way. You'd think a show with a premise so wholesome wouldn't be so damn boring.


 No.928717

>>928716

> trashy A-1 anime

It's a trashy Silver Link anime. I mostly watch it because I liked the web-novel and want to see it animated. Would watch Arifureta this summer and Kumo later for the same reason.


 No.928723

>>928717

That's even worse. SL actually worked on some good anime. I didn't ask why you were watching trash though to be fair. Stop consuming without thinking.

>wanted to see it animated

Christ almighty.


 No.928724

>>928723

> Stop consuming without thinking.

No. Watching every ongoing anime is important for otaku. It gives you bigger base for comparison, helps to discover hidden gems and enhances your taste. Watching every ongoing is almost as important as watching every older anime.


 No.928733

File: 8985b02abc8378a⋯.png (639.75 KB, 640x480, 4:3, Bleach 016 [Zero HP].mkv_s….png)

>>928724

Hahaha. Sure kid. Watching the 7th episode of a LN commercial gives you a bigger base for comparison. You're mentally deficient or what?


 No.928735

File: dcfec2d20a4dc71⋯.png (458.05 KB, 635x640, 127:128, dcfec2d20a4dc71c98f4219037….png)

>>928716

>shit shit shit shit shit

What anime is good?

>>928733

>Bleach

Okay.


 No.928741

>>928735

Why did you reply twice as to pretend to be a different person? Even your reaction images are tarnished with product placement.


 No.928752

>>928673

>You're quite right, anime is produced in 720p,

That's wrong. There are some anime with 1280x720 as native resolution but most anime are produced in "weird" resolutions between 720 and 1080. Which means that the 720 version is a downscale and the 1080 version an upscale.

Anibin (https://anibin.blogspot.com) is a Japanese blog that checks and posts the native resolutions every season.

The master resolutions for some of this season's anime are:

Sewayaki Kitsune no Senko-san : 1440x810

Nande Koko ni Sensei ga!? : 1552x873

Fairy Gone: 1280x720

BokuBen: 1536x864

Kimetsu no Yaiba: 1520x855 (This is the standard native resolution for Ufotable)

Senryuu Shoujo: 1552x873

Hitoribocchi: 1440x810

Carol and Tuesday: 1280x720


 No.928783

File: 958af13766a1083⋯.jpg (55.94 KB, 600x737, 600:737, to_what_end.jpg)

>>928752

>all these meme resolutions


 No.928799

>>928783

>to what end

Selling the prdocut it's advertising me thinks. God Eater ended up looking amazing.

>>928741

You don't get it. A girl having Doritos for a haircut is funny because it references pop culture.


 No.928834

File: 8e75c3cd5ca829a⋯.png (731.41 KB, 884x720, 221:180, Untitled.png)

>>928783

Bad studios.

They either maliciously do it to save resources knowing it'll be upscaled to whatever resolution it'll be on television/blu-rays without giving a fuck about delivering an image that's actually the resolution it's sold at, or they're tech illiterate and think scalers are perfect. Either way shit's very fucked indeed.


 No.928849

>>928752

It doesn't make a difference, because the 1080p version is upscaled from the downscaled 720p version every time.

The comercial version will always be 720p, be it on TV or Bluray.


 No.928860

>>928849

Simply false. If you can't tell the difference between 1080 and 720 you're blind.


 No.928862

>>928676

It's a vicious catch 22 with Gentoo. On a system like mine where it'd be desirable, it takes too long to compile. To that end I am running Debian but with a few programs built from source -O3 -march=native so that would be mpv, FFmpeg, and the dependencies libx264, libx265, libopus, fdk-aac, libvpx.

Also I don't know what I'm lookng for on that Shift+I thing. I sometimes use I to determine things about the video file but that's it. Also you just prompted me to check watching the video in RAM, instead of on the HDD since it's slower, and also has to go through LUKS. No dice. Ultimately, GM45, and 1.87GHz dual-core is only sufficient for 720p and I really don't need 1080p at 1280x800.

>>928683

Even 2TB is constraining for me, and that's as big of a drive as fits in my laptop. The 4TB ones are fatter.

>>928783

>meme resolutions

So is this actually a thing?

https://blog.n4o.xyz/blog/memeres/


 No.928870

>>928860

>No proofs

Nice.


 No.928877

>>928741

It's a low quality reaction image with outdated memes because your low quality posting doesn't deserve better.


 No.928878

>>928877

Low quality posting doesn't even deserve a reply. Replying with low quality posting on low quality posting you just increase an amount of low quality posting.


 No.928900

>>928752

So, why? Is it really that much harder to just draw shit at 4k raw and downscale it from there? Are they just tech illiterate?

What about movies, what resolution are those done in? Upscales are going to look like shit with cinema.


 No.928904

>>928716

>How would one (me) go about this?

waifu2x has a commandline version for unix systems which can convert a single picture. If you sequentially convert the frames of a video with ffmpeg to a format it can accept, upscale them, and then reassemble them again with ffmpeg, you can appropriate waifu2x for upscaling a video.

Otherwise there's AviSynth.


 No.928906


 No.928909

>>928900

> Is it really that much harder to just draw shit at 4k

Yes, it's hard. 4k has 9 times bigger area than 720p, so animators would have been forced to draw 9 times more, and there would have been more place for mistakes and QUALITY to happen. It would take a lot longer, because bigger need a lot more details to look good. And a lot of modern drawing software works like shit with 4k, so there is need for more powerful computers and a lot more time to process such video. 4k is meme and almost no one use it for a good reasons.


 No.928911

File: e487ca716d01b29⋯.gif (548.01 KB, 450x338, 225:169, 1507540587131.gif)

>>928870

Prove what? If you can't see the difference in a resolution you're blind? How do you suppose resolutions (higher than) are calculated if the image is 720p?

>>928877

>I was shitposting because I don't like discussion

Ah noted. But can you please go somewhere else then? Because you don't get a free-pass to shitpost.

>>928878

What about a low-quality reply to an otherwise good post in an intention to disqualify a post? What if someone says something that's so blatantly wrong that you can't help but be baffled someone actually goes about and think such? How can you genuinely believe that watching the 7th episode of a LN commercial with shit-tier production values is important for otaku? It's some fallacy shit and you know it. You don't need to watch it in order to find hidden gems or enhance your taste. The 7th episode has NOTHING to do with enhancing your taste of finding hidden gems. To me it just comes of as a mentally-ill no-drop faggot that doesn't know watching something doesn't make you special. Now is my post low-quality just because I used a swear?

I made the assumption because I know they exist and behave similarly. What's wrong with that? What's wrong with telling people to think before they consume? Imagine not dropping a shitty LN commercial. Imagine having standards. Imagine using a no-true scotsman to validate your inability to drop shows. I care about anime. It's not just another timesink for me, just like posting here, I genuinely enjoy it.


 No.928914

>>928911

>What's wrong with telling people

Egocentrism. You are not qualified to command other people to change their ways of enjoying their life, nor your advices are required. Your whining just derails the conversation and doesn't serve any positive purpose, so your posts are of poor quality. Even if you can't understand something, it doesn't mean it's wrong or bad, especially when it concerns other people. Every person is different and sees things from the different perspective.


 No.928918

>>928914

I'm not telling anyone to change their ways. Your entire post is based on a false premise.

>stop whining

If you can't attack my argument, feel free to continue attacking me as a person. But know that conversations work with arguments. And there is definitely nothing subjective about having standards.


 No.928921

>>928911

>talks shit about several anime

>when asked to prove his superior taste, conveniently ignores it

>claims samefag on everyone who calls him out on being a faggot

>still doesn't see how his posts could be considered poor quality

>he wrote wordswordswords, so they must be good posts


 No.928922


 No.928923

>>928911

>t. Tryhard retard


 No.928924

>>928909

>4k has 9 times bigger area than 720p, so animators would have been forced to draw 9 times more

You know about the drawing -> filming process, I hope. Doing everything digitally is the problem here.

>>928918

Not that I'm condemning his shitposting but

>Every person is different and sees things from the different perspective.

Holy shit, go back to reddit, you insufferable relativist intellectual.


 No.929071

>>928922

Any repo differences?


 No.929087

>>929071

Check them both out and run diff? Code-wise there's not a lot. Are those repos even related to each other, as in you can pull/push between them?


 No.929269

File: 6d28c15aa7c4c86⋯.jpg (58.89 KB, 960x720, 4:3, 1378146732533.jpg)

I am a faggot that uses Daum Potplayer with LAV Filters. I just use Spline scaling 3 times the input the resolution for any additional clarity. I also have a Plex Media Server for watching my backlog on tv.

I use XmediaRecode for encoding video, MKVToolnix to mux or extract assets, and sometimes Sony Vegas for resyncing audio.


 No.929730

>>929269

> Potplayer

Boo

> Plex

Then use Shoko Anime with it


 No.929741

>>929269

No, you're just a normal idiot.


 No.930064

File: 52d03b2d1cb2db3⋯.png (29.9 KB, 800x600, 4:3, __yorigami_shion_touhou_dr….png)

>>928909

I hope the advent of 4K television will make anime so costly that animators go back to drawing on resolution-independent cels to save money.

It'll never happen but a man can dream.


 No.930067

>>930064

Upscaling will probably get good enough that they don't have to by that point.


 No.930068

>>930064

> advent of 4K television

It's unlikely to happen. Television popularity is dropping and going to 4k needs a lot of money to enhance bandwidth. Let's say one 4k TV channel needs 40 Mb/c bandwidth, so 100 channels would need 4 Gb/c bandwidth to every household just for TV. Infrastructure can't handle this. TV would sooner die and be replaced by youtube, than go to 4K.


 No.930072

>>930068

>Let's say one 4k TV channel needs 40 Mb/c bandwidth, so 100 channels would need 4 Gb/c bandwidth to every household just for TV. Infrastructure can't handle this.

<what is Nipponese fibre infrastructure

<what is AV1

<implying every TV channel will adopt 4K right away

<implying 4K "TV" won't be entirely IP-based

Sadly NN scalers will probably take over at that point and NN inbetween animation too considering how Anime Studios seem to be running out of animators.


 No.930078

>>930067

What the fuck is upscaling? I keep seeing people say "upscale" as if it's a good thing all over the world. As if stretching an image 4x over will suddenly change anything, other than just being a lower resolution image. Upscaling.


 No.930079

>>930072

Yeah, I was talking about IPTV. It basically transmits every channel at the same time, and you choose, which one to recieve. As far as I can find, Japanese infrastructure is overloaded and can poorly handle even the modern data streams.


 No.930084

>>930078

> I keep seeing people say "upscale" as if it's a good thing

It isn't really a good thing. However, it's almost unavoidable. In the ideal world, you would get the same video resolution, as your display has, but the world isn't ideal. If you watch old 480p dvd-rip using 1080p display, you either watch watch it in the window, or upscale it to your display resolution. Your video-player does it in real time. Modern upscaling technics are becoming better and better, so you can get a decent image quality even with inferior source.


 No.930087

>>930068

This. In my country people still have to pay for 1080p instead of 720p channels.


 No.930115

>>930064

>I hope the advent of 4K television will make anime so costly that animators already making literal slave wages make negative income just so studios can cover the price of cels and film

<saves money

You're kind of retarded aren't you? And 4K will never become the resolution to animate at. We might see anime mastered in 4k down the line (as in drawn in a lower resolution, but the final output is upscaled to 4k), but not as the standard animated res. Learn how the industry works before saying dumb shit.

>>930078

Depending on the type of scaling used, upscaling can work. This is especially true if you're upscaling from HD (720/1080p) to UHD (4K), which can look a lot better than say upscaling from SD (480/576p) to HD, which usually tends to look like shit.

>>930087

>In my country people still have to pay for 1080i instead of 720p channels.

Fixed. Outside of some pay per view and sports events in the US and Japan, there are no 1080p channels anywhere.


 No.930118

>>930078

>>930084

>>930115

Yeah the technical term is "super-resolution" and it involves a lot of Machine Learning to get optimized.


 No.930119

>>930118

> it involves a lot of Machine Learning

Not necessary. It's done without machine learning more often than with it.


 No.930120

>>930119

Traditional Upscaling and "Content-aware Upscaling" are very different as one is more noticeable than the other


 No.930122

>>930120

It depends on the source. As a person who wrote a Master's Degree on topic of image processing, I can say that in some cases machine learning behaves worse than linear interpolation algorithms.


 No.930142

Machine learning is unintellectual and lazy. Besides, the results are more often than not crap if you're not dealing with simplistic raw numbers in a sheet, but with graphics or movement.


 No.930143

File: a26b8d5e80cd1b8⋯.png (1.19 MB, 1058x2382, 529:1191, slide.png)

>>930122

> that in some cases machine learning behaves worse than linear interpolation algorithms

In what cases that would that be true? Would like to know as a CS student.

>>930142

> the results are more often than not crap if you're not dealing with simplistic raw numbers in a sheet

You can actually see the difference if you compare between the original vs downscale-standardUpscale vs downscale-NeuralUpscale

Tell me then, what do you think, are happening to Neural Upscaling when compared to the original and standard upscaling?


 No.930149

>>930143

Not one of the anons you're replying to.

Neural network based upscaling isn't upscaling at all. It creates a new image based on the old data.

You can't make an image contain more data than it already has by upscaling. Where is that data supposed to come from?

I the case of Neural network based upscaling a neural network gets trained with sample data. When used it simply compares and adjust things to textures from the sample data.

I think you'd get a different image if you downscaled it again.

And Waifu2x denoise shits over all little details nearly just as much as over the artifacts.


 No.930152

>>930149

> Neural network based upscaling isn't upscaling at all. It creates a new image based on the old data

They call it "Super-resolution"

> You can't make an image contain more data than it already has by upscaling. Where is that data supposed to come from?

In a sense it guesses how it should look given the entropy provided. There is no increase of entropy

Some have already suggested using NN as a type of dictionary-esque lossy compression algorithm but I don't think it is practical

> I the case of Neural network based upscaling a neural network gets trained with sample data. When used it simply compares and adjust things to textures from the sample data

Which is why anime/manga image data is used for Waifu2x, and other Super-resolution tasks uses different datasets

> I think you'd get a different image if you downscaled it again

I will be very close to the original downscale though

> And Waifu2x denoise shits over all little details nearly just as much as over the artifacts

That happens due to the fact that you have to train it to know what is a detail and what is an artifact (in upscale-only mode)


 No.930156

>>930143

> In what cases that would that be true?

When image has a lot of noise and artefacts, what isn't that uncommon in older video. Neural Networks tend to make noise and artefacts sharper and more visible in comparison to linear interpolation.


 No.930157

>>930156

So Linear for SD to 720p, NN for 720p to 1080p/4K, or 1080p to 4K?


 No.930159

>>930149

>You can't make an image contain more data than it already has by upscaling. Where is that data supposed to come from?

Some of the data is provided by the algorithm itself in the form of constraints on what it expects the final image to look like.

Take the recent imaging of a black hole's event horizon for instance. The measurements they started with were equivalent to a very low resolution image with huge gaps in fourier space. There were an infinite number of possible high-resolution images which matched their measurements, but most of those were just random noise. They knew roughly what they were looking for (some arrangement of bright blobs on a dark background), so they were able to massively reduce the search space and generate a series of likely images.

Similarly, waifu2x knows that the real image is probably made up from a number of relatively flat color regions bordered by lines, plus a few shadows and highlights.

It's a bit like how you can beat the diffraction limit in an optical system if you know that you're looking at a point-source (or several point sources).


 No.930165

>>930157

Yeah, but you may want to experiment with bicubic, bilinear and Jinc for SD, depending on how shitty the video is. There is no silver bullet, sadly, and each video needs player to be specificlly configured. If you are not as autistic as I am, just leave it at Jinc for SD.


 No.930192

>>930165

So Jinc is the best option when non-autistic?


 No.930384

File: f30c39b0ba9ca58⋯.png (471.02 KB, 924x768, 77:64, 8261c8027183a07cd994ad6d55….png)

>>930067

Nope, most studios still use bilinear to upscale. They're unqualified for their own profession.

This will remain true even after machine learning scalers are amazing. In fact, machine learning scalers are already amazing, and they're not using them, nor is scaling the only thing studios mess up. Far from it.

>>930143

>In what cases that would that be true?

Since linear interpolation is just a mathematical algorithm which finds a middle point between 2 points (and the algorithm gets extrapolated for more points), it would be true when this assumption is correct. Which for anything but artificial images specially crafted for this effect, such as a digitally drawn perfect checkerboard pattern at a perfect 90 degree angle from the screen, or a flat gradient, it will not be true.

Even with these examples, depending on whether you're upscaling or downscaling, and how exactly the pattern is drawn, linear can create an incorrect scale.

In the real world, there's exactly 0 reason you'd use linear scaling for anything, and the only reason it's still in use is because people are often unqualified for what they do, it was the first algorithm and is also the simplest, and poor GPU and graphics API design (bilinear scaling is guaranteed to be hardware accelerated by a GPU, any other form of scaling is guaranteed to not be).

The other anon is just a shitposter.

>>930149

>I think you'd get a different image if you downscaled it again.

You would, and you also would with "oldschool" scalers, because they aren't necessarily reversible, and even if they are, it's almost guaranteed that the result was rounded and information was lost anyway.

>>930157

You use whatever filter is mathematically proven to be a closer approximation of what the image should be and/or approximates better what our human eyes and brain can definitely imagine it should look like.

These are 2 aspects we hope to achieve with different scaling techniques.

If by NN you mean nearest neighbor, you should almost never use it, the only use case I can come up on the spot right now is when you play an old game and you want to keep it pixelated as it was, as whatever scaler you use will add a certain amount of blur which will be farther from how things should be as just using no scaler at all will get you.

You should never under any circumstances use linear, instead you should use in order of preferrence:

>AI scaler (nnedi3, waifu2x, KrigBilateral, etc) > sinc function approximation scaler with EWA (lanczos3 family) >sinc function approximation scaler without EWA (also lanczos3) > keys cubes scaler with EWA (mitchell, catmull-rom, bicubic, etc) > keys cubes scaler with no EWA (mitchell, catmull-rom, etc)

EWA=Elliptical Weighted Averaging. Normally scalers consider pixels equally when scaling, no matter how relevant they are for the scaled pixel to be created. This is simply wrong, and you can give more weight to the ones who overlap more with the final pixel, EWA is an implementation of that. It's a little heavier since it's doing more, but the image quality gains are there.

If your hardware is too weak for any of those, then you have no choice but to go bilinear. If your GPU's bilinear hardware somehow can't do the scaling you need, which will never happen unless your hardware is too weak to decode the video you're playing to begin with, then you have no choice but to go nearest neighbor.


 No.930386

File: 5d7926ee5152e66⋯.gif (709.15 KB, 500x537, 500:537, se-lain-g8.gif)

For MPV users, here's my thoroughly researched config.

You will have to read mpv.conf and input.conf and adapt some options to your OS and hardware, because a lot of it is hardware and OS-specific. You may also have to tone it down if your hardware isn't better or equal to what is described on mpv.conf .

It's not perfect either, but it's a very well researched config.

https://megaupload.nz/radcz1t7n4/mpvconfig_7z

https://anonfile.com/v2d6z8tfnb/mpvconfig_7z


 No.930391

>>930386

>gtx 780

>i7-4770k

It would be hard to find something weaker than that in 2019.


 No.930753

>>930386

Linux config vs windows config?

>>930391

Any ideas on a better config?


 No.930758

>>930391

>he doesn't watch anime on a Radeon X1950+Athlon 64 rig from 2006 running NetBSD


 No.930775

File: 86aa5d9b4d310a3⋯.jpg (128.37 KB, 1024x927, 1024:927, Daijoubuu.jpg)

>Want to overlock my AMD phenom x4 955 Black edition

>Raise it from 3.2 GHz to 3.5 GHz

>Crashes after 4 hours

>Raise HT link frequency to 2400mhz to keep up with my 1600mhz ram

>Bump the voltage a little

>Wont boot and the BIOS has to be reset

Meanwhile I keep reading that people just bump their X4 to 4.0 Ghz no problem.


 No.930776

>>930775

Temperature. Upgrade to a newer Zen in the future.


 No.930929

>>930775

Did you increase voltage sufficiently? My Phenom II X4 965 used to run perfectly well at 3.8 GHz, 1.4V until it burned to crisps 2 years ago. Though, I must say that this CPU is extremely trashy and even core i3 8100 is better. You should conceder an upgrade.


 No.932748

any good torrent clients?


 No.932752


 No.932753

>>932748

qBittorrent


 No.932764

>>932748

I use deluge on linux. I'm seeding over 300 torrents and it's only using 150 MB of memory, so it's pretty efficient. It can be remote controlled so it works well on headless servers too.


 No.932770

>>932764

It has shitty GUI and works like shit on Windows, though. qBittorrent is a way to go, if you use good OS.


 No.932856

White men use Transmission while waiting for Synapse to mature.


 No.933243

>>932770

>>932764

>>932856

what about a torrent client that can handle symlinks? (because right now I am trying to write a script to connect torrents with Hydrus)

(Hydrus imports a file to the system into a flat folder, the previous torrenting folder will need to be symlinks for easy connection and access)


 No.933244

>>933243

If you're on a good OS then symlinks should work relatively seamlessly.

There's a deluge plugin which lets you run an arbitrary command when a torrent completes. You could use that to automatically move the file to your Hydrus folder and leave a symlink in its place.


 No.933838

>>933244

link please to the plugin. Also which OS is good?




[Return][Go to top][Catalog][Nerve Center][Cancer][Post a Reply]
Delete Post [ ]
[]
[ / / / / / / / / / / / / / ] [ dir / agatha2 / b2 / choroy / dempart / thebestb / vg / vichan / webcams ]