2009-04-22, 10:44 | Link #121 |
Pioneer in Fansub 2.0
Join Date: Aug 2007
|
Always post the raw unfiltered source (as in cropped & IVTC'd but nothing else) / lossless source screenshot for comparison purposes as well.
But frankly with the filtering you mentioned earlier they all look quite bad. Why can't people just set for preserving the original quality instead of fucking around with (over)sharpening and whatnot?
__________________
|
2009-04-22, 11:18 | Link #122 | |
x264 Developer
Join Date: Feb 2008
|
Quote:
The difference between 21 and 24 is three. The difference between 18 and 21 is three. The difference between 15 and 18 is three. Since the scale is logarithmic, all of those differences are the same, percentage-wise. Raising the quantizer by 6, no matter what the quantizer currently is, is roughly equal to cutting the bitrate by half. A linear quantizer, for comparison, is one where doubling the quantizer cuts the bitrate in half. In other words, 2 to 4 is the same as 4 to 8 is the same as 8 to 16 is the same as 16 to 32 (etc). Also, most of the 175MB screenshots look fine to me, to be honest. |
|
2009-04-22, 11:33 | Link #123 | |
Senior Member
Author
Join Date: Jul 2007
Location: Virginia Tech
|
Quote:
In other words, a^x=b, given a and b, find x. So, basically what you were saying. XD
__________________
|
|
2009-04-22, 12:05 | Link #125 |
Pioneer in Fansub 2.0
Join Date: Aug 2007
|
More like it proves that you might not need more than 175MB for HD shows if you introduce enough artefacts just by filtering alone, which is what Kristen has done in this case.
Seriously, if you're introducing very notable banding and other nasty artefacts that wasn't originally there just with the filtering, you're doing something very wrong.
__________________
|
2009-04-22, 12:22 | Link #127 | |
Senior Member
Author
Join Date: Jul 2007
Location: Virginia Tech
|
Quote:
It didn't add artifacts to the encode. What happened was that the banding present got accentuated by the grain removal. If anything, wouldn't artifacts DECREASE compressibility? It's more information to process each frame.
__________________
|
|
2009-04-22, 12:45 | Link #128 | |
Translator, Producer
Join Date: Nov 2003
Location: Tokyo, Japan
Age: 44
|
Quote:
18+3 = 21 18+6=24 but that doesn't mean that 21 is 1/9'th worse than 18 or that 24 is 1/3 worse than 18. That's what people normally assume, they'd think it was like temperature. Look, I've tutored high school physics in the past, so I know how to simplify the concept of logarithmic scale to people. If you say that "any difference in 3 is the same, percentage-wise," I think that will just confuse anyone who was confused to begin with . It's like when explaining the ricter scale... The difference in energy between magnitude 7 quake and mag. 6 is a lot bigger than the difference between a 4 and a 5. I'm talking about absolute bitrate, not percentage bitrate.
__________________
|
|
2009-04-22, 12:48 | Link #129 | |
Banned
Join Date: Nov 2003
Location: Hamburg
Age: 54
|
Quote:
The 175 pictures already have visible ringing in still frames. Which means that once you have motion, this ringing creates unstable lines and exactly the kind of mosquito noise which is the visible indicator for bitrate starvation. Also, with AQ activated at strength 1, you can get certain kinds of artifacts around edges when you're short on bits, so if you DO select this option - a very valuable and helpful one, no need to explode now, Dark Shikari - make sure you have sufficient bitrate. FWIW, I don't quite see what the nagging about your filtering is about. The sequence is a bit unusual, normally you clean first and sharpen afterwards. But that's all. |
|
2009-04-22, 12:48 | Link #130 |
Multilingual Girly Geek
Fansubber
Join Date: Apr 2009
|
The 175 MB images do have some extra artifacts, but other than that, the difference is minimal. And, minor banding aside, they all look good enough to my inexperienced eyes.
Regardless, I think we might be missing the point here. It doesn't matter if my connection is good enough to easily hit several Mbps on an average day. The point is that I'm lucky to hit a 10-15 Kbps download speed for any semi-popular anime series a few months after it's completed due to a lack of seeders. So filesize does matter, especially if it means I'll get to download an episode in only half an afternoon instead of the next day. And it'd be nice to be able to fit a series in less than 4-5 DVDs when I pass it on to my friends. On top of that, please remember that this forum represents a small minority of what I'd call 'hard-core fans'. Opinions here might be a bit skewed in favor of quality, since most people out there will hardly notice a couple more pixels here and there in a high action sequence. |
2009-04-22, 13:01 | Link #131 |
Banned
Join Date: Nov 2003
Location: Hamburg
Age: 54
|
The fans out there HAVE decided on this issue. The vast majority favors hi-resolution hi-bitrate releases - just look at download numbers. It's the "small filesize" group which gets ignored.
And sorry, if you have difficulties downloading fansubs via torrents within a few hours, you must either be several weeks late or have some really bad ISP. |
2009-04-22, 13:07 | Link #132 |
Senior Member
Author
Join Date: Jul 2007
Location: Virginia Tech
|
Went with the first 17 seconds, since there's little contrast at the starting frames, and mkvmerge split it into 17 second chunks instead of 5 like I asked it to. >.<
700 MB: http://www.megaupload.com/?d=6N2M1E70 350 MB: http://www.megaupload.com/?d=416UTPHD 233 MB: http://www.megaupload.com/?d=WKGJKDWF 175 MB: http://www.megaupload.com/?d=CTT96G1I Honestly, way I see it is the 175 MB is too little, 233 MB is the balance between quality/filesize, 350 MB is the quality encode, 700 MB is just a waste of bitrate.
__________________
|
2009-04-22, 13:14 | Link #133 | |
Senior Member
|
I wasn't saying 175MB is identical to 700MB, I was saying that 175MB is perfectly acceptable to most people
Mentar, the sad truth of the matter is, the majority of fansub viewers watch them on crap quality streams like Veoh and Youtube and Crunchyroll, so clearly the majority doesn't give a crap :/ Quote:
|
|
2009-04-22, 13:26 | Link #134 | ||
Banned
Join Date: Nov 2003
Location: Hamburg
Age: 54
|
Quote:
The quality/filesize question in the fansub watcher community is answered. Quote:
|
||
2009-04-22, 13:31 | Link #135 |
Banned
Join Date: Nov 2003
Location: Hamburg
Age: 54
|
Agreed. The ringing caused by 175 disturbs my eye quite a bit, the 233 version is okay with only occasional blimps which attract my eye, and 350 is really slick. If I had the free choice, it would probably be around 300 megs. But both 233 or 350 (I'd go 340 for DVD sizing) would be perfectly valid choices in my book.
|
2009-04-22, 13:32 | Link #136 |
Senior Member
|
I love how definitively you speak on things, Mentar. You should go into politics
On topic though, as long as you and other fansub groups continue to release a hardsubbed xvid SD version for legacy users like myself (SD TV with old set-top box), I don't care if you need a RAID array to store the HD releases |
2009-04-22, 13:45 | Link #137 | ||
Banned
Join Date: Nov 2003
Location: Hamburg
Age: 54
|
Quote:
Okay, touché. Your point. Still, we had this kind of discussion for years. And the download numbers have clearly developed away from smallres avi to hires mkv. Once they were 5:1 avi:mkv, now they are between 1:2 - 1:3 avi:mkv. I think that does count as an answer. Quote:
Number of hires releases that can be stored on this disk: ~4.500 Tears shed by me for inducing storage costs to poor anime fans: Guess. |
||
2009-04-22, 14:22 | Link #139 |
Pretentious moe scholar
Join Date: Oct 2006
Location: Vancouver, Canada
Age: 37
|
File size is seldom an issue for me unless as long as I can download off IRC at 600 kb/s. When some weird routing issue is causing my downloads to average 30kb/s as has been happening the past few days I do tend to notice quite a bit though. I wish I knew who to complain to to get that issue fixed.
I actually still use DVD-R as my primary means of storing stuff because there's not way in hell I'd trust my collection to a single hard drive and I can't afford a redundant setup. That said, am I correct in thinking that this CRF thing produces smaller file sizes for the same quality level? In that case, I'd have to say just use that, since it should technically mean that I should be able to fit more on a single disc. The only time file size bugs me is when groups do stupid things like making a 26 episode series 70MB too large to fit on a single DVD-R. And most of those groups use constant files sizes - what the hell. |
2009-04-22, 14:26 | Link #140 |
Senior Member
|
I know it's an old concept, but some groups did that to discourage archiving in the hopes you may actually buy the DVDs when it gets licensed. I know that's why Hikari no Kiseki always used 180 megs (or whatever it was). Obviously it's more of a symbolic gesture than anything.
|
|
|