This is an encode of the release made by reza: https://nyaa.si/view/1660009
The main motivation for this release was to reduce the filesize (obviously). So I denoised the source quite hard sacrificing some detail in the process. I tried to recreate the grain with av1 grain synthesis which turned out a little less grainy then the original. I want to point out that I didn't bother to generate a film grain table for every single frame and rather chose to use an average estimation for all episodes because it would have taken forever to compute.
You can find some comparisons here: https://slow.pics/c/Fs1GwQxW
Other Seasons: https://nyaa.si/user/sagiriz?f=0&c=0_0&q=inuyasha
```
Video
ID : 1
Format : AV1
Format/Info : AOMedia Video 1
Format profile : Main@L4.0
Codec ID : V_AV1
Duration : 24 min 27 s
Bit rate : 1 657 kb/s
Width : 1 440 pixels
Height : 1 080 pixels
Display aspect ratio : 4:3
Frame rate mode : Constant
Frame rate : 23.976 (24000/1001) FPS
Color space : YUV
Chroma subsampling : 4:2:0 (Type 0)
Bit depth : 10 bits
Bits/(Pixel*Frame) : 0.044
Stream size : 290 MiB (87%)
Default : No
Forced : No
Color range : Limited
Color primaries : BT.709
Transfer characteristics : BT.709
Matrix coefficients : BT.709
Audio #1
ID : 2
Format : Opus
Codec ID : A_OPUS
Duration : 24 min 27 s
Bit rate : 129 kb/s
Channel(s) : 2 channels
Channel layout : L R
Sampling rate : 48.0 kHz
Frame rate : 50.000 FPS (960 SPF)
Bit depth : 16 bits
Compression mode : Lossy
Stream size : 22.6 MiB (7%)
Title : English
Writing library : Lavc61.3.100 libopus
Language : English
Default : Yes
Forced : No
Audio #2
ID : 3
Format : Opus
Codec ID : A_OPUS
Duration : 24 min 27 s
Bit rate : 124 kb/s
Channel(s) : 2 channels
Channel layout : L R
Sampling rate : 48.0 kHz
Frame rate : 50.000 FPS (960 SPF)
Bit depth : 16 bits
Compression mode : Lossy
Stream size : 21.7 MiB (6%)
Title : Japanese
Writing library : Lavc61.3.100 libopus
Language : Japanese
Default : No
Forced : No
```
i used svtav1 with
```
--crf 23 --preset 5 --fgs-table grain.txt --color-primaries 1 --transfer-characteristics 1 --matrix-coefficients 1 --chroma-sample-position 1 --keyint 10s --scd 1 --enable-overlays 1 --tune 2 --enable-qm 1 --qm-min 0 --enable-tf 1 --input-depth 10
```
and `core.knlm.KNLMeansCL(clip, d=3, a=2, h=2.6)` for denoising
I tried some other denoising options and combinations that did preserve a little more detail but all of them were kind of to slow for my taste
(i have no idea what i'm doing)
Nice efficiency and the grain table is pretty close. At `d=3`, everything is going to be slow. Denoising chroma as well at lower values may have been helpful.
S02 should be ready in the next couple hours
I tried also denoising chroma but it looked exactly the same or even a little worse
`d=3` isn't even that slow at least I'm still cpu limited on my hardware
Comments - 6
NekoTrix
sagiriz (uploader)
NekoTrix
Ironclad
laurensz
sagiriz (uploader)