New codec

[gee]

Hello,

i was just wondering why subbers use divx and xvid, but not better codecs like VP6, RV10, WMV, and soon the H264 from altene / nero.

The quality is way better for the same size.

Some people started using matroska, but we also need better codecs no ?

Thanks

[SirCanealot]

VP6, RV10 and WMV9 are NOT better than XVid.
I've heard H.264 beats XVid, but as it only current fits into a mp4 container, it's not very pratical for use right now, imo.
Either way, which ever side of the coin you're on (eg: admitted XVid fanboy like me; though that's only because it's worked so well for me), XVid is still perfectly confirgerable and highly compedative as a codec - I doubt you'll see people stop using it for a LONG while yet.

Although I admit I've been meaning to try out Nero's H.264 for the past few days, but with no VFW support, I'm too lazy ^^

[bayoab]

Originally Posted by gee

Hello,

i was just wondering why subbers use divx and xvid, but not better codecs like VP6, RV10, WMV, and soon the H264 from altene / nero.

The quality is way better for the same size.

Some people started using matroska, but we also need better codecs no ?

Thanks

These codecs are not better. Lets start going down the list.

http://www.doom9.org/codecs-203-5.htm#conclusion
(yes, its old, but it still holds up.)
Vp6: "VP6 didn't excel in any domain and could definitely use some polish when it comes to animated features." (Instant loss.)
RV10, proprietary player only basically, limited support for older computers. Yes, it is very good at compressing animation, but the lack of any 3rd party player kills that advantage.

WMV is far inferior for animation. It is also proprietary IIRC. (You need WMP9 series to play back WMP9 enocded stuff, wmp6.4 has issues). There are other legacy issues with it.

MKV is not better than AVI for a normal sub. Infact, it is worse because it is not backwards compatible.

Xvid: Good, all around, No backwards compatibility issues if used right. Decent compression.

[complich8]

the questions to ask are "what do I get from changing" "what do I have to do to change" and "is it compatible with everyone that I care about".

Xvid is the norm. Gives good quality, highly configurable for the encoder, lots of customization, can be done in a compatibility-conscious way. Easy, ready support for linux, mac, windows, modded xbox, relatively easy time making it compatible with divx hardware decoders. Old computers can decode it well enough, new computers can decode it even better. Xvid has a lot to offer, including the fact that pretty much everyone that wants to play it can play it.

The original poster of this thread proposes changing to any of 4 "better" codecs. But ... until there's support for all of those on all of those platforms, that's your first loss in switching.

What do you gain? Arguably better video quality. At the cost of less compatibility with the applications that people are using (eg: SirCanealot's VFW comment), less compatibiltiy with players, hassle and headache getting the viewers to be able to view the stuff at all, loss of cross-platform support, loss of the most popular ways to play stuff.

The mere existence of something you believe is better doesn't necessarily warrant the effort and difficulty and sacrifice needed to migrate to it. The community as a whole didn't adopt xvid until it was clear that the benefits of using it outweighed the cost of moving to it, and the same will (and probably should) be the case for any other format, codec, or standard.

[Mentar]

One should note that this doom9 comparison test was conducted by someone with very little experiences especially in anime encoding. Or, to put it bluntly, the author had no clue what he was doing. Whoever uses DivX for anime encoding using psychovisual enhancements can't be taken seriously. Same for several other issues. For example his conclusion for wmv9 is way off - it's actually very good for anime encoding, if it's utilized correctly (the picture must be presharpened to offset the internal smoothing of the codec).

Most top anime encoders I know have been ignoring doom9's "reviews" in regards to anime encoding ever since, or at least read them with a big deal of scepticism.

Also, mkv is strictly superior to avi. Even if you use the "dumb mode" and hardcode the subs into the video, accepting the inherent quality loss, it still offers more audio stream support than avi, along with lots of other advantages. The ignorance of fans should not be taken as an acceptable excuse to forfeit progress. Otherwise we'll stagnate on AVI forever. Installing splitter software which is freely available in any common OS is not too much to ask.

[SirCanealot]

Yeah, WMV9 can work extremely well for anime at times.
Some WMV9 raws I've seen have looked miles better than the DivX5 raws.

The only thing I have against WMV9 is the power required to decode it:
DivX/XVid encode = 500mhz max
WMV9 encode = 090348549058205820385mhz mininum
Really, any time I see anything going over half way on my 2.1 while JUST decoding video gets me worried. This also means it takes MORE time to re-compress a WMV9 raw to XVid :P

[bayoab]

Originally Posted by Mentar

One should note that this doom9 comparison test was conducted by someone with very little experiences especially in anime encoding. Or, to put it bluntly, the author had no clue what he was doing.

Experience... you do not need experience in anime encoding. Encoding is encoding is encoding. You just need to know handle your source video. There is no special settings, there is nothing special about it. Divx3, XVID, DIVX, etc all allow you to just plug your source in no matter what it is, and it is smart enough to handle the rest.

For example his conclusion for wmv9 is way off - it's actually very good for anime encoding, if it's utilized correctly (the picture must be presharpened to offset the internal smoothing of the codec).

This, in itself is a flaw. Codecs should not require anything special. The doom9 test was simply "Plug it in, what comes out." It was not a test of "Lets see, if you apply 500 filters, you can get the perfect image with this codec because THAT IS TRUE OF ALL THE CODECS."

Originally Posted by Mentar

Also, mkv is strictly superior to avi. Even if you use the "dumb mode" and hardcode the subs into the video, accepting the inherent quality loss, it still offers more audio stream support than avi, along with lots of other advantages.

1) How often is a fansubber going to use more than 1 audio stream?
2) That is not an advantage, that is a difference in capabilities. AVI was not made to support multiple audio streams. Also, MKV needs a special player to take advantage of these features. It does not properly work in WMP6.4.

I am still missing a true example of how it is superior for a fansubber. For dvd rippers, you have an arguement.

The ignorance of fans should not be taken as an acceptable excuse to forfeit progress. Otherwise we'll stagnate on AVI forever. Installing splitter software which is freely available in any common OS is not too much to ask.

Backwards compatibility is not much to ask either. Yet the past 3 versions have not been backward compatible. (I still use the original Ffdshow final release and have 0 issues with any properly encoded (no hacks) avi.)

I personally have to watch mkv video's in GRAPHEDIT (yes.. start laughing all who know what that is) that utilize anything more than the avi standard.

[K_R]

People who laugh don't realise graphedit is for cool people.

[megumidk]

I can imagine that most encodes in divx/xvid because is the most widely used. And as to the ones you mentions are better, that is often a matter of opinions and more importently, how they are used.

But do we really need better quality fansubs ?
After all, most anime will be released in superiour quality on DVD, so better quality in fansubs seems to be more for the benefits of those who do not purchase the DVD's then in the interest of the fans.

[SirCanealot]

Originally Posted by bayoab

Experience... you do not need experience in anime encoding. Encoding is encoding is encoding. You just need to know handle your source video.

Then tell me under what optional conditions should QPel be used with anime encoding. What matrices best suit certain anime sources. How XVid responds to certain types of filtering.
Thankfully, XVid and its defaults certainly are smart enough to give good results coming from decent quality video (I've heard otherwise for DivX5's defaults and anime), but this issue is still far from saying XVid is putting in and getting out: if someone knows how to tweak XVid for their source, then so much better will the quality be.

And trust me, encoding live-action and encoding anime and two VERY different kettles of fish. Although, as you said, that test was simply a "put it in and see what comes out" test(very low bitrate too, wasn't it like 600kbs?); which is why it shouldn't be used as any sort of serious reference to how codecs perform under real conditions (eg, properly filtered and ran at much higher bitrates).

This, in itself is a flaw. Codecs should not require anything special.

In life, EVERYTHING requires special treatment to get the best out of something. How do you protect against XVid's like to block certain scenes (usually dark or murky ones) to high hell? I don't even know what's up with DivX5, but I'm sure it has its fair share of problems that can be somewhat offset by using video tools...

1) How often is a fansubber going to use more than 1 audio stream?

Never, but I can say for a fact, if more raw-cappers started using mkv and OGG audio, and more fansubbers did the same, we'd have a hell of a lot less bad-sounding releases about :P

And I'm sorry, but backwards compatibility? .mkv still doesn't work in Windows95/98? Too bad. If you've got a mototcar from the 1950s, you can't expect to be able to put the latest motor engine in it without some serious work.

And yes, I agree with K_R completely and utterly for once: Graphedit is a badman program. Even though I've only rarely used it, I can see how sick the program is. If there was one program I wished I'd used more in my life, it'd certainly be Graphedit. There's no limit to how badass graphedit is, in my opinion. One encoder I know (back in the dark days of muxing mkv) said he was scared of Graphedit. I had to say "pffft, you're not a badman; this is how yo run this shit, bro".

But do we really need better quality fansubs ?

As has been explained before, most (decent) encoders have huge egos that state they have to work almost 24/7 in making anime look the best possible. Phrases like "lol, this video filter makes teh anime look slightly better, and only adds 4 hours onto the encode!" and "wowz0rs, encodergasm!!" are fairly common if you hang around a half decent encoder. Actually, it's kinda scary at times.
But it's like, learning a new technique, apply it, tweaking it, then having something come out looking better than the original raw ever would (well. Always from a certain point of view) is like... winning the olympics, or something. It's like a private and personal victory only for you, and the few others that also share the eyes to see what you've done with the video...
At least that's how I feel, perhaps in somewhat random words ^^

[Sakaki-]

Originally Posted by gee

Hello,

i was just wondering why subbers use divx and xvid, but not better codecs like VP6, RV10, WMV, and soon the H264 from altene / nero.

The quality is way better for the same size.

Some people started using matroska, but we also need better codecs no ?

Thanks

VP6 Made for streaming video thru the internet and not for making files to distro.

RV10 Definitly not better then XVID on anime.

WMV Built in filtering that screws with your source in ways you as a encoder dont want.

H264 http://www.mpegla.com/news/n_03-11-17_avc.html <-- says it all

Take Care
Sakaki-

[Mentar]

Originally Posted by bayoab

Experience... you do not need experience in anime encoding. Encoding is encoding is encoding. You just need to know handle your source video. There is no special settings, there is nothing special about it. Divx3, XVID, DIVX, etc all allow you to just plug your source in no matter what it is, and it is smart enough to handle the rest.

That's exactly what I mean by lacking experience. While I'll admit that _in theory_ you are correct, and that a codec should perform good under all circumstances with all sorts of video footage, _in reality_ things don't work like this. When it comes to encoding video, what makes or breaks a video is how it's filtered BEFORE it's passed to the codec. And the special quirks of the chosen codec dictate what kind of filtering you have to apply.

If you honestly think that Divx3-5 or XviD don't care about the kind of footage you give them, you still have a long way to go.

This, in itself is a flaw. Codecs should not require anything special. The doom9 test was simply "Plug it in, what comes out." It was not a test of "Lets see, if you apply 500 filters, you can get the perfect image with this codec because THAT IS TRUE OF ALL THE CODECS."

Again - in theory, yes. In real life, no. All mentioned codecs have distinct strengths and weaknesses. EXPERIENCED encoders are aware of that and choose both codec and pre-codec filtering appropriately.

The main issue I had with the doom9 article was that it came to entirely wrong conclusions. Like for example that WMV9 and DivX5 were unsuited for anime, while that's actually the strong area for both codecs. And EXPERIENCED encoders know not to use the artifact-inducing psychovisual enhancements for video. This codec test was NOT conducted by a knowledgeable anime encoder, and his conclusions were somewhere between misleading and plain-out wrong.

I'd argue for a more realistic approach: Codecs are to be tested based on their potential. In other words, what's the best possible quality achievable with competent filtering. This is where there's _definitely_ differences between the codecs, and the choice should be made based on THIS assessment.

1) How often is a fansubber going to use more than 1 audio stream?
2) That is not an advantage, that is a difference in capabilities. AVI was not made to support multiple audio streams. Also, MKV needs a special player to take advantage of these features. It does not properly work in WMP6.4.

1) You misunderstood me in this part, I meant to say that MKV supports a broader choice of audio codecs. MP3 is bad and inefficient, but still the choice for most fansub releases. I hope that this will be changed soon.

2) MKV does not require any special player, virtually any player nowadays supports proper multiaudio playback. The only exception are the horrible WMPs. Again, installing the matroska splitter and virtually any media player has the same complexity as installing the DivX codec had back then a few years ago. This isn't too much to ask, IMHO.

I am still missing a true example of how it is superior for a fansubber. For dvd rippers, you have an arguement.

Just the fastest list:

- More supported advanced audio codecs, saving bitrate
- Less muxing overhead. Yes, sometimes muxing an avi with the same audio file is smaller in mkv than in native avi. I kid you not.
- Soft subtitle support. Once mkv makes its way, fansubbers can skip on the quality-lossy video recompression to hardsub their tracks.
- Multilanguage subtitle support
- All other mkv container advantages, like chapters etc.

Backwards compatibility is not much to ask either. Yet the past 3 versions have not been backward compatible. (I still use the original Ffdshow final release and have 0 issues with any properly encoded (no hacks) avi.)

That's a capacity of the chosen codec, MKV however is a container. Install the splitter and that's it.

I personally have to watch mkv video's in GRAPHEDIT (yes.. start laughing all who know what that is) that utilize anything more than the avi standard.

Erhm ... no offense intended, but why are you doing this?

[subcool]

i rather liked DivX 3.11a (using SBC encoding) ^^;
looks like XviD will be leading the fansubber encodes for quite a while.

btw, if ya'll want quality, go release a filtered huffy =P

[Mentar]

Originally Posted by Sakaki-

VP6 Made for streaming video thru the internet and not for making files to distro.

A qualified "yes", since that's its origin, but it works quite okay for distro files too.

RV10 Definitly not better then XVID on anime.

I haven't tried RV10 yet, but based on feedback from some qualified encoder friends it's actually pretty good. However it won't make it due to the fact that it's both proprietary AND not available everywhere.

WMV Built in filtering that screws with your source in ways you as a encoder dont want.

But which you can tame if you know how to do it. WMV9 is one of the least-understood codecs there is, and yet it works surprisingly well on several source types. Whoever fought very hard to beat certain WMV9 encodes with DivX or XviD will know that it can be very difficult sometimes.

H264 http://www.mpegla.com/news/n_03-11-17_avc.html <-- says it all

Does it? I'll venture a guess that H.264 will become the codec of choice over the next few years. While it's still in development, it already has some impressive results to show. The licence situation is IMHO fairly irrelevant, since freely available filters (ffdshow/libav) are able to play it back.

[mattheus123]

I dunno what you all are talking about. If you dont have full proof evidence to back it up then dont talk. I already tested vp6, divx, and xvid.
The results in order is xvid -> vp6 -> divx.
note: i used divx 5.1.1 in this test. Also test used base thru multipass for each codec.

Divx leaves a trail of dirt -_- .
vp6 doesnt leave as much dirt
xvid rules!

specs: all encode done progessively without any psychovisual enhancement.
each is done through standard method(i.e multipass-first -> multipass-second)

xvid seems to win among this test.

About H.264, maybe this codec will be prefer from many.

[lamer_de]

- More supported advanced audio codecs, saving bitrate
AAC works with avi, ditto DTS. Don't know of any other advanced audio codecs. Maybe ogg (which indeed does not work with avi), but that seems to be optimised for low bitrates.

- Less muxing overhead. Yes, sometimes muxing an avi with the same audio file is smaller in mkv than in native avi. I kid you not.
While I have not tested this myself extensively, savings should probably range in the 100KBs for a file with 1 video and 1 audio stream of 24mins length. Neglectable, if you ask me.

- Soft subtitle support. Once mkv makes its way, fansubbers can skip on the quality-lossy video recompression to hardsub their tracks.
Possible in avi as well with AviMux-gui and vsfilter. Also, nobody wants to distro 350MB files for 20mins of a/v. Or 150MB noisy ones. The main point of encodig is not to add subs, but to deliver pleasing looking video at reasonable filesizes. Those points do not apply to most of the raws, so you'd still have to re-encode.

- Multilanguage subtitle support
Neglectable. For most languages with sufficient interest, subbing scenes already do exist (.fr, .br, .de, .it to name a few)

- All other mkv container advantages, like chapters etc.
I will give you that, although I don't see the point. The only chapter markings that make sense are OP/Eyecatch/ED and well, I never felt the urge to jump to those points directly.

If you want to use another container, use .mp4 :P

CU,
lamer_de

[K_R]

Originally Posted by SirCanealot

Then tell me under what optional conditions should QPel be used with anime encoding. What matrices best suit certain anime sources. How XVid responds to certain types of filtering.

None (possible compatibility problems). None (possible compatibility problems). Apparently the same as with no filtering (as far as I am aware, xvid doesn't have the ability to test if a filter is being used and change its behaviour accordingly).

Originally Posted by SirCanealot

this issue is still far from saying XVid is putting in and getting out: ... In life, EVERYTHING requires special treatment to get the best out of something.

I just put it in and press encode. And my filter chain looks like this:
avisource()
textsub()
textsub()
textsub()

Originally Posted by SirCanealot

Phrases like "lol, this video filter makes teh anime look slightly better, and only adds 4 hours onto the encode!" and "wowz0rs, encodergasm!!" are fairly common if you hang around a half decent encoder. Actually, it's kinda scary at times.

I took out all the useless filters (ie. all filters), so I could encode in 20mins, because, quite franky, encoding shits me. I think my encodes look pretty good though. I recently released ebichu 21-22, so you can judge for yourself, my encoding ability (with no filters!)

Encoding is such an easy thing to do, I'm surprised encoders can actually trick other group members into thinking it requires some special mojo.

Note to encoders: learn a new skill, that's vital to a group (like translating), before you get the boot! ;)

--
p.s. I remember a time when encoding with xvid was something radicals did. People would complain to me about not using the standard divx... How times have changed.

[Mentar]

Originally Posted by lamer_de

- More supported advanced audio codecs, saving bitrate
AAC works with avi, ditto DTS. Don't know of any other advanced audio codecs. Maybe ogg (which indeed does not work with avi), but that seems to be optimised for low bitrates.

And this is exactly why it would be particularly suited for fansubs.

- Less muxing overhead. Yes, sometimes muxing an avi with the same audio file is smaller in mkv than in native avi. I kid you not.
While I have not tested this myself extensively, savings should probably range in the 100KBs for a file with 1 video and 1 audio stream of 24mins length. Neglectable, if you ask me.

It adds up with the audio. Only by itself, it's no big issue, but all things combined it starts to be one.

- Soft subtitle support. Once mkv makes its way, fansubbers can skip on the quality-lossy video recompression to hardsub their tracks.
Possible in avi as well with AviMux-gui and vsfilter. Also, nobody wants to distro 350MB files for 20mins of a/v. Or 150MB noisy ones. The main point of encodig is not to add subs, but to deliver pleasing looking video at reasonable filesizes. Those points do not apply to most of the raws, so you'd still have to re-encode.

I beg to differ. Most good quality raws nowadays are decent enough, and adding the subtitles for another hardsub doesn't improve the quality, it lessens it due to the recompression. In the fewest cases the fansubbers manage to "improve" the video (then the chosen raw must have been horrible), they only add the subs and reduce the size, usually to 175 megs.

If the use of mkv would be more widespread, the RAW providers could shoot for the right size by themselves, so that the fansubbers would only have to mux in the subtitles. Which is how it should be done (from a purist's standpoint), every recompress reduces the achievable overall quality.

- Multilanguage subtitle supportNeglectable. For most languages with sufficient interest, subbing scenes already do exist (.fr, .br, .de, .it to name a few)

Exactly. If mkv was used, it would be much easier to simply add the extra subs. Which is something which slowly spreads in the DVD world. Insisting on avi forces them to make their own full releases with full encodes.

- All other mkv container advantages, like chapters etc.
I will give you that, although I don't see the point. The only chapter markings that make sense are OP/Eyecatch/ED and well, I never felt the urge to jump to those points directly.

I did, and I do. Skipping the OP is a convenient thing to do.

If you want to use another container, use .mp4 :P

Heh... hardly ^_^

Sure you can argue "hey, with this and this hack you can do that with avi, too". But why should we do that, if a technically better and sound solution is already available? Simply because the fan community might take a while to get their playback systems set up? I disagree with letting laziness prevent technical progress. The DVD world pretty much has moved on to MKV already and the way things are developing, it's only a matter of time until the fansub world follows suit. It has already begun.

[Sides]

Originally Posted by Mentar

If the use of mkv would be more widespread, the RAW providers could shoot for the right size by themselves, so that the fansubbers would only have to mux in the subtitles. Which is how it should be done (from a purist's standpoint), every recompress reduces the achievable overall quality.
...
...
...
Exactly. If mkv was used, it would be much easier to simply add the extra subs. Which is something which slowly spreads in the DVD world. Insisting on avi forces them to make their own full releases with full encodes.

here is the thing
if a subtile file is only mux, it can be demux easily.
And fansubbers rarely/don't give away their scripts, for few reasons.

besides: compatibility over quality

[Sakaki-]

K_R then you haven't seen a good encoder at work and your not going to be counted as one cause that is not even considered an effort.

And who said it was hard to do a basic encode?

Mentar you wont see anything but mp3 in fansubs till the raw capers change format.
Stuff like this you say make me wonder if you ever encoded tv source yourself.

Take Care
Sakaki-