[Lazy] Suzume (2022) v2 (BD 4k SDR) (2160p) [6297C976].mkv

2023-10-16 02:22 UTC
Discord: witchymary
File size:
20.3 GiB
Info hash:
This release is for >1080p SDR screens. If you have an HDR screen, download an HDR version of the movie. [Here's a comparison between this release and the official SDR Blu-Ray](https://slow.pics/c/PnNCuj2g). Subs received very minor tweaks. ___ | | | |------------------------|---------------------------------| | Translation | R1 | | Encoding | motbob | | Timing | witchymary |

File list

  • [Lazy] Suzume (2022) v2 (BD 4k SDR) (2160p) [6297C976].mkv (20.3 GiB)
Still some minor differences compared to [mpv tonemapped HDR -> SDR](https://nyaa.si/view/1726070#com-5) but new v2 encode looks a whole lot better. Added grain bloat though. Also still missing chapters.
Thank you gamers
> Added grain bloat though I'm not aware of any grain being added on my end.
> I’m not aware of any grain being added on my end. Hmm. Thought perhaps you might have been re-graining but it might just be the encode size preserving more micro-detail / noise. (Beatrice slightly de-noised, AnimeLand could have as well.) Besides some ever so slight color tinting differences, I like the encode best now. Doesn't have AnimeLand's pixel bleeding at the top/bottom or Beatrice's AA (anti-ringing?) warping / destruction. Subs could have used some more work like in NeoDESU.
In any case, we are waiting for Blu-ray editions from other countries. The Japanese edition of both discs has a lot of problems. I have an idea to eliminate upscale artifacts from 1920x804, as they write here. https://hatsuyukidate-my.sharepoint.com/:x:/g/personal/s_knyu_littlebakas_moe/EfcrYpVGAfRPl1oNjlig0CsBC-AJ-_QOo_ghM35E1ocJCA?rtime=-sBIK92620g But at the same time, it seems to me that this information is not correct. In any case, to check you need to wait for another publication with a higher bitrate.
[@Jensen](https://nyaa.si/view/1729838#com-5) Start by being able to successfully find the upscaler used for simple native 720p shows before attempting anything like that. Dehaloing/deringing after incorrect descales is destructive, it attempts to hide the issue with the incorrect rescale (albeit leaving artifacts behind many of the times). You'll find that a correct rescale will eliminate upscaling artifacts including halos/rings and upscale-induced aliasing, at the expense of other minor ones, caused by doubling and then downsampling. You're better off just doing everything else _but_ descaling if you can't figure out a correct one.
The problem is that a neural network was used here, and not traditional upscale kernels.