RMVB: Why not use it?

[TheFluff]

Whatever size you may prefer when downloading anime doesn't matter much - the answer is not RV40 and/or RMVB.

No, not for ULTRA COMPRESSION GET stuff either. I made a 25 MB x264+HE-AAC test encode earlier today, mostly because I was bored (similiar encodes has been done before), and to see how bad it would be (props to checkers for inspiration and advice). Surprisingly, it didn't turn out half as horrible as I thought it'd be. I was planning to make a similiar RealVideo encode to see how it'd compare, but RealProducer Basic doesn't let you edit any encoding settings yourself - you can only use its preset profiles, and the only one that went as low as I wanted it was a 15fps CBR streaming profile. The lowest VBR profile was a 350kbps one (including audio). I didn't feel like testing this, since the my version was about half of that (112kbps video, 32kbps audio).

The results of my testing can be seen here: http://www.cccp-project.net/beta/tes...64+HE-AAC).mkv
The karaoke and the group logo is hardsubbed, mostly because I wanted a "realistic" scenario. Removing all hardsubs would probably help quality a fair bit - they look pretty bad as it is, despite me having applied extra antialiasing to them. Setting a lower chroma QP offset would probably help some as well (I overused it, leading to evil blurring).

EDIT: removed karaoke, did some tweaking of settings.

Some stats:
Video resolution: 432x320 (displayed as 432x324)
Target video bitrate: 114 kbps (x264 rev523, Sharktooth)
Target audio bitrate: 32 kbps (Nero commandline AAC encoder, May 01 build)
Source: 640x480 DTV XviD raw (ASKA)
Subtitles: ASS, SRT (softsubbed, fonts attached)

Maybe someone more skilled with RealProducer than I would care to try to encode something similiar and compare?

[jpwong]

Originally Posted by Farix

Are you so sure? There plenty of people who are stuck with dialup that would wish to download fansubs, but the current file sizes makes it extremely inconvenient, taking a day or two to dedicated downloading just to get one episode. So they just don't even bother unless they are really dedicated. That is why I see the trend of larger file sizes with AVC encodings when the codex can achieve smaller files sizes for the same quality as current XviD encodings being rather detrimental to the community as a whole. The larger encoding will put those encodings well out of each of those who do take the time to down fansubs over dialup.

Ok, well, it is a problem for people stuck on dial up, but I think it would be a hard sell to get fansub groups to encode ~50MB videos while trying to maintain their elite (ego driven sometimes) quality. I only know one group that offers lowres dial up friendly encodes (~40MB lowres RM and ~110MB h264 and ~170MB XVID of the same show), and they aren't exactly mainstream subbing.

[Sylf]

Originally Posted by TheFluff

but RealProducer Basic doesn't let you edit any encoding settings yourself - you can only use its preset profiles

I've only played around with Helix Producer in the past, but it was almost the same. No, you can't edit the profiles using the software. But in Helix Producer, all the presets were defined in an XML file. I don't see how it could be different in RealProducer as well. With the XML file, you can tweak the exact bitrate of both audio and video, decide which codec to use, cbr/vbr, framerate, etc, etc. Reading through all the documentation to actually understand all the codes you need to put for the exact result was the most painful part of using Helix Producer. (And Helix website wasn't easiest site to navigate, which didn't help me at all)

[Starks]

Originally Posted by TheFluff

Whatever size you may prefer when downloading anime doesn't matter much - the answer is not RV40 and/or RMVB.

No, not for ULTRA COMPRESSION GET stuff either. I made a 25 MB x264+HE-AAC test encode earlier today, mostly because I was bored (similiar encodes has been done before), and to see how bad it would be (props to checkers for inspiration and advice). Surprisingly, it didn't turn out half as horrible as I thought it'd be. I was planning to make a similiar RealVideo encode to see how it'd compare, but RealProducer Basic doesn't let you edit any encoding settings yourself - you can only use its preset profiles, and the only one that went as low as I wanted it was a 15fps CBR streaming profile. The lowest VBR profile was a 350kbps one (including audio). I didn't feel like testing this, since the my version was about half of that (112kbps video, 32kbps audio).

The results of my testing can be seen here: http://www.cccp-project.net/beta/tes...64+HE-AAC).mkv
The karaoke and the group logo is hardsubbed, mostly because I wanted a "realistic" scenario. Removing all hardsubs would probably help quality a fair bit - they look pretty bad as it is, despite me having applied extra antialiasing to them. Setting a lower chroma QP offset would probably help some as well (I overused it, leading to evil blurring).

Some stats:
Video resolution: 432x320 (displayed as 432x324)
Target video bitrate: 112 kbps (x264 rev523, Sharktooth)
Target audio bitrate: 32 kbps (Nero commandline AAC encoder, May 01 build)
Source: 640x480 DTV XviD raw (ASKA)
Subtitles: ASS, SRT (softsubbed, fonts attached)

Maybe someone more skilled with RealProducer than I would care to try to encode something similiar and compare?

(Damn, I'm good at striking up conversation...)

I certainly have the sanity to try... But my skill is pretty much shit.

[Farix]

Originally Posted by jpwong

Ok, well, it is a problem for people stuck on dial up, but I think it would be a hard sell to get fansub groups to encode ~50MB videos while trying to maintain their elite (ego driven sometimes) quality. I only know one group that offers lowres dial up friendly encodes (~40MB lowres RM and ~110MB h264 and ~170MB XVID of the same show), and they aren't exactly mainstream subbing.

I'm not talking about reducing AVC encodings ~100MB, though it could probably be done with little lose of quality. Just that AVC encoding shouldn't be bigger then their XviD counterparts--which has been the trend lately with AVC and which I see as being completely counterproductive and harmful to fansubing in general. And using the excuse, "well downloaders are on broadband anyways," just strikes of arrogance that really irritates a dialup slave like myself.

[DryFire]

Originally Posted by Farix

Just that AVC encoding shouldn't be bigger then their XviD counterparts--which has been the trend lately with AVC and which I see as being completely counterproductive and harmful to fansubing in general.

I only know of one group that releases the h264 version in a larger filesize then the xvid version when both are at the same res, and the h264 version is the HQ version while the xvid version is simply a comapatbility 26 eps/dvd thing. The remainder use the same or a smaller filesize.

I don't see how higher quality is harmful to fansubbing. A lot of groups are choosing to go the 233mb xvid round instead of 175mb's h264, I would consider that counterproductive.

[ntsan]

Some of the fansub groups in China release both avi and rmvb format, but avi only got less than 1/4 of the downloads than rmvb

I think why chinese prefer rmvb is because it is a step up from rm (which was widely used before in china for virtually everything from episodes to movies) just like how we prefer avi from the past to now

Preference > other

Like how most people still use IE even though FF/Opera is much better

[jpwong]

Originally Posted by Farix

And using the excuse, "well downloaders are on broadband anyways," just strikes of arrogance that really irritates a dialup slave like myself.

Most groups making larger h264 encodes I've seen that have h264 larger than xvid have been opting to use larger res videos, sometimes significantly larger res video (2x the pixil count of the xvid version in at least one case I think), most cases I've seen with identical sized video, the h264 encode is either the same size or smaller than the xvid counterpart.

And given the number of downloads you see on some torrents, it's not hard to imagine that most shows are being consumed by people on broadband. Digisubbing hasn't ever IMO really been dialup friendly, the filesizes vs download times have always seemed to have turned out products more suitable for higher end connections. Since groups are supposedly making these videos with themselves as the primary consumer (I suppose that's highly debatable these days), it's natural to assume that if the whole group is on high speed, they likely aren't going to put a huge consideration into low filesize, whereas a group made up primarily of dialup users will likely put filesize before ultimate quality.

[checkers]

Well, there have been some interesting posts in this thread with interesting claims that I wanted to test. The argument comes down to how they compare, using the best tools available. It's no good having some mythical Real encoder that can do 25mb encoding if you don't actually have access to it.
One RV10 encoder I did have acess to however is Real Anime. Real Anime 4 includes encoding support for RV10 video encoding (not audio, and I doubt it would be as good as AAC anyway) at custom bitrates. So, I got out the trusty unlicenced raw and got to work.

The source:
The file was Honey_and_Clover_-_07_-_We_look_for_a_miracle_[Raw].avi, a 720p 120fps source which was noise free and with minimal banding. It was run through cfr2tc and the raw avi used as input for the AVS script:
avisource("raw.avi")
bilinearresize(480,272)
Dup(threshold=0.8,blend=true)
The dup() setting is a bit high, but nothing to seriously worry about for a test.
Both file have the same 35kbits AAC-HEv2 included and subs (OCRed from solar's release with zero spellchecking :P)
The x264 Encode:
--pass # --bitrate 146 --stats "avs.stats" --keyint 999 --ref 16 --mixed-refs --no-fast-pskip --bframes 16 --b-pyramid --b-rdo --bime --weightb --direct auto --filter 1,1 --subme 7 --trellis 2 --analyse all --8x8dct --me umh --progress --no-psnr --output NUL "avs.avs"
Nothing special here - options were reasonably normal. Special interest is the ...'interesting' GOP size and that's about it. Download it here
The RV10 Encode:
Done through Real Anime 4. Bitrate 146, high quality, 2 pass vbr, "RV10 DropDupe" was unchecked. Despite this, the encoder produced:

• an oversize file (the video was 20.8mb as opposed to 20mb or so, which is a good 4%)
• a file with dropped frames.

Even though I told it not to, RV still seems to drop some frames all the time, producing it's own vfr output. The problem with this is that I needed to mux the file with my own timecodes to produce a vfr output! Trying this resulted in a video that played back in fast forward, and trying without the original timecodes resulted in a video that started running slow by 25% during the main section of the episode. In other words, I could not get any useable output with RV10, but here's a link to the file. The only problem is the video play speed, you can still judge quality. Download it here.

I was going to post screenshots, but they don't really reveal any difference - you'll have to watch them to see. If you are too lazy and trust my judgment, the h264 file is far better. The real file drops frames and looks jerky, and is far blockier, even during still scenes.

[physics223]

All I can say is wow. And you're only 18. Damn you for being smart with encoding!

I'd still use RMVB (not to encode, mind you) but to watch when all else fails.

[IRJustman]

Snagged the file and played it with MPC (I have RealPlayer Alternative installed for the Real codecs and (in this case, not used) splitters). Once the show started after the opener, it was evident the sound and picture were already out of sync. I can only imagine the wrangling you must've gone through to try to get things even remotely right.

Laying all that aside, the video's decent if you don't care too much about quality. In its defense, it might be useful to preview something to see if you want to commit to a lot of disk space and bandwidth usage for a series, but that's about it. I certainly wouldn't use something like this for real distribution.

Besides, if I wanted to do something like this, I'd use VP3 (the basis for Theora)+HE-AAC+NSV, basically the basis for Nullsoft TV. All you'd need is a recent copy of Winamp and you're good to go. All of the standards are open (Real's is not) and have been implemented on multiple platforms.

--Ian.

[TheFluff]

I suppose I should post my x264 commandline for reference purposes.

x264.exe --pass 2 --stats ayakashi08.264stat --bitrate 114 --bframes 16 --b-bias 20 --filter 3:2 --subme 7 --merange 24 --keyint 600 --ref 16 --trellis 2 --ratetol 5.0 --chroma-qp-offset 3 --pbratio 1.2 --scenecut 35 --me umh --analyse all --8x8dct --b-pyramid --b-rdo --bime --weightb --mixed-refs --no-fast-pskip --threads 2 --thread-input --progress --output compressiontestv2.mkv final.avs

[checkers]

Yes, in my opinion 25mb is too low for decent quality files. It starts getting interesting around 50mb, and DryFire mentioned 86mb as being 52eps/DVD, also an interesting size. As for your VP3 idea, I'd star far away for something this small. Quality would be worse than xvid, which would be worse than both of these. If you wanted something truly open, you would be stuck with theora/vorbis/ogm though, MPEG4 spec is technically open, but definately not free .

I guess to get back to the OPs question before this thread is even further derailed, the two reasons why I would hate to use RV:

• Not natively supported by anything.
• As well as lack of support, there is also a distinct lack of control even in the programs that do use it. RA4 cannot fully disable the vfr'ness, which means for any anime that requires it (and that's more than you think if you don't want jerky sections) cannot properly be encoded with it.
• The only legal decoder for it is Real Player. Real Alternative is illegal. Probably not something the Highly Ethical Fansubbing Scene wants to do (sorry, couldn't resist )
• Lower quality than h264.

The one reason you would use RV seems to be the decoding power required.

• h264 with ffdshow: 20%
• h264 with CoreAVC: 15%
• RV with Real Alternative: 10%

Not bad hey? That said, don't draw too many conclusions. This encode was done at a tiny resolution, using a tiny bitrate, and using the highest possible x264 settings (not what would normally be used). I'd bet that at normal 640x480 or 704x480 resolutions the difference would be noticeably less.

IMO, the lack of tools, the fact that you will negate the increase in quality from xvid because of it's stupid frame dropping and the fact you can only legally play it through Real Player (which ranks right up there with VLC for a great GUI and QT/BSPlayer for flawless compatibility) means although it may have been a great codec, it isnt .

[TheFluff]

Judging from the few tests I did, RV didn't seem much faster (if indeed faster at all) than h264 at sane resolutions and somewhat normal settings. I didn't test it with timecodec though, and now I've already deleted all my RV samples.

BTW, checkers' encode is much nicer than mine, despite me using the following AVS to kill all semblance of detail:

undot()
aa()
unhalo(rad=3)
temporalsoften(5,6,8,15,2)
hqdn3d(6)
deen("a2d",5,2,3)
dup(threshold=0.5,copy=true)
Crop(2,6,-2,-6)

textsub("D:/ayakashi/08/aya08 - ts-vfr.ass")

aa(edge=true)

bilinearresize(432,320)

I blame his nicer and more compressible source. >_>

[complich8]

First of all, the broadband prevalence discussion earlier in this thread is totally irrelevant to the question at hand. People on dialup may struggle to download a few ultra-compressed episodes of various stuff, but ~21MB/hr download speed, even with "dialup-friendly" episodes, they're not going to get very much done. I don't think it's reasonable to even consider catering to that demographic, regardless of their misfortunes. Broadband penetration is on the rise everywhere -- in the US, it's at like 68% now (source). Game companies (think STEAM), software companies (think windowsupdate), and the latest and greatest on the net (youtube and google videos, flikr streams, espn's front page) aren't really catering to dialup anymore. If they aren't, why should we? Fansubs are all about video, and no other video service gives a second thought to dialup users.

I personally don't like getting or using new formats until they are stabilized enough and well-supported enough to work on the 3 major platforms (windows, mac, linux) without jumping through major hoops to get there. I used to be more open to that, but I'm not really interested in fighting with a container or a format or finding out that my computer's just a little too slow to handle whatever the latest and greatest is, or that my version of some media splitter or another is missing a feature that the latest release depends on.

Technical heterogeneity can be a good thing, but for the English-speaking fansub-viewing audience (or, perhaps, for consumers as a whole), it absolutely isn't.

For the technophile, playing with the new formats, whatever they are, is a good time in and of itself. I remember watching in amazement as my roommate babbled on incessantly about how awesome those first HD-resolution h.264 movie trailers looked. He wouldn't shut up about it! Even though his monitor is only big enough to display roughly half of a 1080p video...

But most people aren't that guy, and aren't really interested in spending time and effort keeping up with the latest and greatest in codecs and players. They want simple, they want it to work, they don't care if you managed to make a great looking 720p x264 encode in 130 megs while curing bovine melanoma. All they care is that it works without them having to futz around with it.

All of this applies to rmvb too. As you make the environment more heterogeneous, you make people have to spend more time getting set up to watch stuff, which means spending less time actually watching it.

This view shows up pretty clearly in real download numbers: take a look at download counts or torrent populations for anything that's released in both x264 and xvid (here's some examples). Notice the trends: x264, despite being heralded as the clearly technically superior solution, is downloaded somewhere between 1/3 and 1/7 as much as xvid, whenever there's a choice.

When I download a conventional, plain-jane xvid-mp3-avi encode of something, I KNOW that I'll be able to play it, in any player I want, on any platform I want, with no problems at all. When I download an h.264-aac-mkv, I know I'll be able to play it in vlc and maybe, maybe not in other players on my platform, depending on what's in the package. When I see an rmvb available for download, I know it probably isn't going to play nicely on anything I've got installed right now, and whatever it is is probably not going to be worth the annoyance of figuring out what I actually need to play it.

Encoders like to choose codecs for compelling technical reasons. But consumers don't care, they want good enough and easy to use. XviD gives them that, because it's familiar. RMVB gives them the same thing in China, because, again, it's familiar, since everyone uses it. Maybe one offers compelling technical superiority, but compelling technical superiority does NOT mean compelling improvements in the overall experience.

This is the same reason that blu-ray and hd-dvd are going to have a lot of adoption problems: dvd is good enough (better than most people's tvs and stereos are capable of showing off) and is pretty easy to use. HD-DVD and Blu-Ray are both dramatically better quality, but again most people don't even have the setups to take advantage of the HD video or 5.1+ audio, so there's not a lot of perceived value that the technically superior offering brings to the table.

In other words, arguing codecs on their technical merits belongs to enthusiasts, developers, and video specialists, but if you're trying to better serve the general public, forget about all that crap and make your decisions about codecs based on what most people are willing and able to play.

[Soluzar]

Originally Posted by Complich8

In other words, arguing codecs on their technical merits belongs to enthusiasts, developers, and video specialists, but if you're trying to better serve the general public, forget about all that crap and make your decisions about codecs based on what most people are willing and able to play.

That works if fansubbing is a consumer-focused endeavor. A lot of fansubbers see it as being anything but.

[Maceart]

Wasn't fansubbing originally intended to get as many people acquainted to anime as possible? If so, then using the easiest to run video codecs is the right choice.

[CelesAurivern]

Originally Posted by Maceart

Wasn't fansubbing originally intended to get as many people acquainted to anime as possible? If so, then using the easiest to run video codecs is the right choice.

Except almost nothing is the same as it originally was.
And viewers these days have a higher incidence of rabies infections; too much barking at the charitable hands that feed them

[aperson]

What did fansubbers want people to get aquainted with?
The Japanese culture? Then sub regular Japanese TV.
The stories? There are many non-anime shows and they have some pretty good stories as well.
The animation style? Then the end user should see the show in the way it was intended to be seen, without any artifacts (or as little as possible) and with clear audio.

If you do not intend to view anime the way it's meant to be shown then ask yourself: Why did you watch anime in the first place?
Are you being pulled into new shows because of a habit of only watching anime?
Do you want to see animation not targetted to children? Why target animation specifically?
What's so special about seeing it animated? If you want to know what happens as soon as possible, just look at screencaps and the summary.
Do the voices appeal to you? I'm sure regular Japanese TV will have some good voices.

To smudge an original piece of art is insulting to say the least.
A replica painting is acceptable for most people but it does not compare to the beauty of the original.

[deathbygirl]

Off-topic and irrelevant. If you want to argue about the relevance of change throughout fansubs with the passing of time, I'm sure many other fansubbers who reminisce about the olden days would love to debate about how "anime is/was intended to be seen" and "how things were when real fansubbers were around" in another thread. This thread, however, is about how relevant RMVB is (or to be more precise, could be) to the lower-end (i.e. non-broadband) of the fansubbing community, which was actually an interesting discussion before (another) discussion of fansubbers' ethics and beliefs ensued.