Our encode that will be used for our [PAS] v1 (unless I find something wrong with it during QC) with slapped on v0 subs as-is for people who really can't wait a bit longer.
This wasn't QCed and the encode may actually change for [PAS] v1 if something is found during QC, but it's frame exact with our web release so theoretically the subs should just work.There is no eta for the [PAS] v1 which will have much improved typesetting, a couple lines looked over, a couple silly things from the v1 fixed, and probably even additional signs typeset (the diary comes to mind)
Included is every audio track that came on the disc (5.1, 2.0, 2x commentary tracks) with the 5.1 track being the default.
Dual track release. One default track without honorifics, one with. Have your player default to enm to play the honorifics track automatically.
Really nice BD. Clean bicubic b=0 c=1/2 upscale (testing was done and that's the conclusion I came to, at least) from 720p , zero banding but lots of noise. Just descaled, denoised then did a light deband to fix any banding that the noise might've been covering up.
Use ravu or nnedi3 upscaling or something similar with mpv or mpc-hc/be and you'll have a better 1080p image than what's on the disk at a much smaller size.
#pa-subs@irc.rizon.net
Recruiting competent translators to save more anime. Contact KoolKidsK on IRC or KoolKidsK#8146 on discord if interested.
File list
[PAR] I Want to Eat Your Pancreas [BD 720p AAC] [B9EAC7FF].mkv (1.8 GiB)
Thank you for the subs, but there's one pet peeve on mine: You absolutely can't determine the so-called native res with downscaling then upscaling back test. This is easily falsifiable by creating an image with 1920x1080 canvas, drawing a rectangle, then applying the test, You can downscale the native 1080p image down the single digits and upscale it back up, and it'll look the same and so the native resolution from that test will falsely result in the single digits. Yet you know it's 1080p since you made it. A lot of behind-the-scenes interviews and studio production footage have proven native resolution to be 1080p or composite of multiple high res assets than what people have assumed.
For example, Kill-La-Kill is native 1080p and in fact, the digital key frame scans used for the animation that Trigger posted up in their blog during production was a bit greater than 1080p. The reason why it doesn't look that detailed is because the director intentionally thickened then *blurred* all the character lines, as revealed in the footage extra DVD vol 2 in the limited ed BDs by Aniplex So a lot people assumed it was 720p with added grain in post-production.
Another example is JC Staff behind the scene footage interview for Kaito Tenshi Twin Angel, an old show from 2011. People assumed 720p (so all the BD fansub encodes were 720p) yet an interview with JC Staff revealed all characters were animated in 1600x960p resolution. Background res is usually much higher. But again, the director intentionally blurred even the background details simply because he literally doesn't like the sharpness. That was an old show and since then it is safer to assume 1080p than otherwise.
It doesn't matter on which resolution it was animated, since usually it's rendered with 720, 810, 840, 864, 873, etc. Descaling is actually needed to undo cheap upscales from rendered resolution, they can't retrieve actually drawn resolution from BDs. About Kill la Kill... There are mixed contents in episodes. Mostly all scenes actually rendered at 720p, but some others looks like 1080p (I didn't yet tested it).
@npz basically what djatom said. Talked to some other encoding buddies too. It doesn't matter what the assets were animated in--everything was rendered at a set resolution. I used DescaleM which used a mask to detect native 1080p elements and not descale them too. Almost nothing these days is rendered at 1080p. Maybe it's safe to assume assets used are often higher than 720p but everything was still rendered at a different, constant resolution and then upscaled to 1080p.
@npz, if stuff really was 1080p and then simply blurred or whatever, then a descale test would never return a value of 720p. It would probably indicate 1080p as being correct. (of course, it gets more complicated it only part of the screen was a 1080p blur, but...)
If the blurring at the studio was accomplished via downscaling, then the information lost during the downscale is never going to come back. So there's no point in the fact that it was originally at 1080p affecting what we do.
> This is easily falsifiable by creating an image with 1920x1080 canvas, drawing a rectangle, then applying the test, You can downscale the native 1080p image down the single digits and upscale it back up, and it’ll look the same and so the native resolution from that test will falsely result in the single digits.
Except it will not look the same at all, at least not mathematically. [This](http://0x0.st/zPFh.png) is what the getnative graph of such a rectangle on a 1080p canvas (`kagefunc.squaremask(core.std.BlankClip(width=1920, height=1080), 200, 200, 100, 100)`) looks like. No encoder in his right mind would descale something like that.
@npz Animations, backgrounds, scene elements etc. are created in different software using vector graphics, so even 4K won't match such accuracy. Then those raw materials are put together in composing stage of production. This is where they got rasterized in 720p or such resolution and where postprocessing, lights etc. are applied. Complexity and cost of composing raising quickly with resolution — that's why it may seem quite low for 2019. At the end, completed composing project is send to some studio involved in production to render the composing project, add credits, upscale to 1080, author BD and TV, and so on.
>Use ravu or nnedi3 upscaling or something similar with mpv or mpc-hc/be and you’ll have a better 1080p image than what’s on the disk at a much smaller size.
>not even dlss
do you even upscale?
My gripe with 720 encodes is that the credits are nearly always 1080p, but encoders do not mention this quality degradation at all. Same issue here, the credits in both the OP and ending credits are quite blurry compared to 1080p and with loss of details. But other than that I have no complains about this encode, quality is much better than the other translated 1080p encode at 3.9 GB, and while there are some compression artifacts, it is not that noticeable and certainly nothing to complain about at that file size.
Powerful enough to sensitive girls. Its book cover says this is a "heartrending tale of friendship and mortality" in the form of a bittersweet romance. I like this anime as much as I like "A Silent Voice."
@PA-Subs Thank you for your hard work on this.
Who the fuck had the idea that 10k+ lines of CPU-heavy scripts in ass would be a good idea? No wonder my older laptop had problems playing this. This one is literally the same only not so hardware demanding if you want an alternative: https://nyaa.si/view/1132379
Made an account to let others know: Quality and subs in this are great, no need to wait for a newer version imo. Thanks for letting me enjoy such a great movie :)
Just watched this...feels, so many feels.
But since it's comments, i'm sure others watched...did they skip why she dumped the boyfriend? did the MC (not in this movie) ever do...'anything' with her?. just curious
Comments - 31
JavierGc
Astral
RyzakiT
npz
DJATOM
PA-Subs (uploader)
motbob
Lypheo
Yogart
Jensen
Aryma
Kazeyuki
xDev
tfw
LCE
torn
SomaHeir
VoidZero
Gunshi
Onii-Sama
Spiller
mokakeyk
lerusso
Mich-666
SwissCheese
lateon
Marv
onizukaftw
Darudius
nelson
RyzakiT