[Sav1or] Golden Time [1080p][AV1][OPUS][Dual Audio]


Source LinksNyaa (cached) | Torrent Download | Magnet Link (5.145 GB) | NZB
Date Submitted25/08/2023 08:40
Series (!)Golden Time
CommentGolden Time
Tracker(s)
udp://tracker.opentrackr.org:1337/announceS: 6L: 2C: 231U: 657d, 18hr, 28min ago
udp://exodus.desync.com:6969/announce-
udp://tracker.torrent.eu.org:451/announceS: 6L: 1C: 189U: 657d, 18hr, 20min ago
http://nyaa.tracker.wf:7777/announceS: 7L: 1C: 158U: 657d, 18hr, 4min ago
udp://open.stealth.si:80/announceS: 14L: 0C: 704U: 336d, 22hr, 43min ago
Files
919.1 KB
286.7 MB
250.8 MB
212.1 MB
230.9 MB
219.4 MB
227.1 MB
228.9 MB
229.1 MB
188.8 MB
200.9 MB
221.8 MB
215.9 MB
254.5 MB
242.0 MB
240.6 MB
158.8 MB
177.6 MB
237.7 MB
198.8 MB
222.0 MB
197.4 MB
200.3 MB
219.2 MB
206.1 MB
SubtitlesAll Attachments



18 comment(s):
Q 15/09/2023 08:25 — Anonymous: "bIGfAN"
The sound on the Japanese and English audio tracks and the video go out of sync on S01E16 at around the 16:07 mark, along with a massive visual artifact.  Would it be possible to get a corrected/proper version of that episode? Thanks.
16/09/2023 02:10 — Anonymous: "Sav1or"
Apologies for the error, and thanks for the report. We'll start on a fix for an eventual v2 release and if there are any other issues, let us know and we will do our best to sort them out. These video stream glitches with AOM were unknown at the time and so this release predates our recent A/V desync checks as well as many other checks and automations we have implemented since. Who would've thought the bleeding edge could be so risky for amateurs?

Thanks again for the heads-up,
-Sav1or
L 19/09/2023 22:52 *Sav1or
Now known as SavI0r
For those who cannot wait for the v2 release, here is a fix for S01E16 using our latest encoding setup: https://multiup.io/68e90983c4bf84a620f...0c647686df

Updated 21/05/2025

Cheers,
-Sav1or
Q 21/05/2025 05:55 — Anonymous
Does anyone still have this? It looks like the link is dead. Please help! Thank you!
L 21/05/2025 08:53 — Sav1or
Now known as SavI0r
We refreshed the link above. We're also working on a better version using the BD remux from CRUCiBLE. We tested some promising rescale parameters (918p, Mitchell) and we plan on making 2 versions with our latest tool, SWAN. Scene complexity in this show is similar to Sakura Quest so we expect far greater efficiency, especially considering all the encoder improvements since the initial release.

Have a nice day,
- SavI0r
21/05/2025 09:29 — Anonymous
Mitchel and catrom spikes are red herrings
21/05/2025 09:57 — Sav1or
Now known as SavI0r
Thanks for the advice! If you have tested parameters for this show then could you share your findings?

We never mentioned how we tested or for what reason we chose these parameters but we don't claim they are perfect either. It's also very likely this show should not be rescaled based on some of our observations but we're no experts so we'll defer to others whenever possible.

Looking forward to seeing your results,
- SavI0r
21/05/2025 22:20 — Anonymous
And thus, "Mitchel and catrom spikes are red herrings" guy was never heard from ever again.
21/05/2025 22:40 — Anonymous
god forbid someone have a life
21/05/2025 22:44 — Anonymous
I have not experimented with this show myself. However past evidence has shown that these codecs just "look correct" due to their inherent low error (https://docs.google.com/spreadsheets/d/1QsQ4SqPmxgOQWFgZ0NVX7LhitVDZEkey0zrtxA4Onco). I'd also like to add that fake rescalers like setsugen no ao, beatrice (all of it's encoders), kawaiika (all of it's encoders), vcb etc have all used these kernels. But I'll give you the benefit of the doubt since there are no comparisons (the one true way to prove descales) provided in your comment.
21/05/2025 22:46 — Anonymous
It would be easier to have this exchange on discord tbh
22/05/2025 05:08 *Sav1or
Now known as SavI0r
Here ya go! We're already familiar with motbob's kernels sheet as well as the handful of kernels published by JET. We concur that they're quite handy for people encoding seasonal or very recent content but this show is unfortunately not on there. Given their strict standards, they may very well deem this show non-rescalable based on what we've seen. We won't get into details as we're not experts, but a simple rescale of just one resolution and kernel across the entire runtime may not be the right answer to acheive a "real" or perfect rescale. Borrowing your words, nothing "looks correct" to us but we're going with the best we got.

A vast majority of the time, bilinear was the right choice for the content we've dealt with in the past. However, due to the haloing already present and introduced by some kernels, we gave Catrom and Mitchell another look to see if we could avoid these artifacts. Having looked at it again with even more kernels, we may reconsider bilinear over mitchell but we need to test other episodes and make samples for playback to really dial in our decision. We added Dehalo to a couple of the kernels to get idea for how much of the artifacts introduced/worsened with the rescale.

Philosophy time. Feel free to skip to the TLDR.

Regarding "real" versus "fake" rescales, the ends sometimes justify the means. If it improves upon the image without deviating too far or introducing noticeable artifacts then all's well. You and many others may vehemently disagree with that and we'll respect your opinion because ultimately that's what it boils down to, an opinion. We're not striving for absolute perfect content that reflects the ground truth either - we're mini-encoders; bottom-of-the-barrel scum in most eyes! Even worse, we use film-grain synthesis to emulate the original grain/noise so we'll never have a perfectly equivalent encode. So as long as the image is faithful to the original and isn't distracting, we don't mind it. Yes, we've seen what happens when things get taken too far and all this testing is to avoid that.

TLDR: All that is to say, we heed your caution and appreciate the info.

Have a nice day,
- SavI0r
22/05/2025 09:52 — Anonymous
no chroma comp?
22/05/2025 11:47 — Sav1or
Now known as SavI0r
At that point we really recommend just loading it up in vs-preview with the native-res plugin installed and playing with settings there. It's easier, faster, and more detailed than the generated slow.pics. In the slow.pics, it's hard to tell a difference between Bilinear and Mitchell in some cases but in vspreview the difference is a lot clearer. That's likely due to the webp compression but we're not trying to spam slow.pics with hundreds of massive lossless images either.

If you'd like to use the same (mostly) script we did to test it then you can take a look here. If you use VS Code you can collapse all the regions except for "Setup" and play with settings there.
Q 22/05/2025 11:56 — Anonymous
Is there any chance for a v2 of episode 14? At 13:05 I'm seeing big artifacts until 13:13. Thank you.
22/05/2025 12:14 — Sav1or
Now known as SavI0r
Howdy, thanks for the report! We can confirm this issue and it looks like it is a similar case to S01E16 and many other releases that were plagued by issues with L-Smash seeking/chunking. Thankfully, we no longer use that so a new version should avoid the issue.

We have plans to redo the whole show with a better source, new settings, and foreign subtitles if we can find them. It'll be a while until we get to it so apologies, we can't offer a replacement right away. We have the source already so we'll be on it as soon as possible but we're working on several projects at the same time so no estimate either.

Have a nice day,
- SavI0r
22/05/2025 12:46 — Anonymous
Thanks for the reply and the great work, and please take your time!
L 04/06/2025 08:46 — Sav1or
Now known as SavI0r
Thanks for being patient with us. Here are early samples of episodes S01E14-S01E16 from the upcoming release. They are missing some subtitles we plan to add before releasing.

We managed to cut the size down nearly 20% from 5.15gb to 4.16gb while maintaining similar quality if not better. That wouldn't be possible without SWAN. Granted, the audio bitrate has gone from 128K to 96K but the video bitrate makes for the bulk of that decrease. The fonts for the FR subtitles are quite large, so their addition will increase the size by ~200mb.

Enjoy,
- SavI0r
Add new comment
Name:
Comment Type:
Message:
Show formatting tags
<b>bold</b><i>italic</i><u>underline</u><s>strikethrough</s>
<code>code</code><sub>subscript</sub><sup>superscript</sup><spoiler>spoiler</spoiler>
<big>big</big><small>small</small><quote>quote</quote><a href="https://animetosho.org/">link</a>
Please be aware of the following before commenting:
  • Anime Tosho provides a mirror of torrents and is not the source. Please understand that uploaders/submitters may not read comments here, so you should check the Source Links section near the top of this page if you wish to contact them
  • Expired links do NOT get reuploaded as files are deleted after we process them
  • Uploads are NOT instant, so please wait for them to be processed. The entire uploading process is done by a bot which does NOT read any comments, so asking for links will have no effect on when they show up
  • Critical comments are welcome, however note that statements such as "X is crap" or "Y sucks" are NOT criticisms. Please provide reasoning with critical comments to be informative, helpful and allow for debate.
  • Personal attacks or insults are not welcome here. Such comments may be removed entirely and commenters may be temporarily banned.
Image Verification:

Our squiggly text game where the aim is to copy the image into the textbox. All characters are upper case, and there are no zeros (0) and ones (1) in the above image. Apparently bots aren't as good as humans at this game.
beta
Anime DDL+NZB mirror
Current Time: 12/07/2025 13:53



About/FAQs

Discord