The H.264/AVC/MPEG4 discussion thread

[Farix]

2) What do you see as obstacles (if any) to the adoption of H.264 and/or modern containers?

Speaking as a leecher I think the biggest reason that AVC hasn't been widely adopted by other leechers has been in how fansubbers had used AVC. Since funsubbers felt that AVC, along with the perceived wide adoption of broadband and larger hard drives, freed them from the constraints of file size, they focused more on high quality encodings. Inevitably, this meant that the AVC encodings were larger then their XviD cousins. However, except for a small minority, file size does matter and those large files gave them the false impression that AVC wasn't as efficient as XviD. I'm also sure that the change from AVI to MP4/MPV also hampered progress in switching, but I don't think it was an insurmountable one.

Most fansubbers have learned from their previous mistakes about quality since then, but I believe that they are now a day late and a dollar short. The damage has already been done and unless fansubbers drop XviD in mass, progress in "leechers" switching to the AVC encodings will continue to be slow.

[Quarkboy]

Originally Posted by Farix

Speaking as a leecher I think the biggest reason that AVC hasn't been widely adopted by other leechers has been in how fansubbers had used AVC. Since funsubbers felt that AVC, along with the perceived wide adoption of broadband and larger hard drives, freed them from the constraints of file size, they focused more on high quality encodings. Inevitably, this meant that the AVC encodings were larger then their XviD cousins. However, except for a small minority, file size does matter and those large files gave them the false impression that AVC wasn't as efficient as XviD. I'm also sure that the change from AVI to MP4/MPV also hampered progress in switching, but I don't think it was an insurmountable one.

Most fansubbers have learned from their previous mistakes about quality since then, but I believe that they are now a day late and a dollar short. The damage has already been done and unless fansubbers drop XviD in mass, progress in "leechers" switching to the AVC encodings will continue to be slow.

Let's not be too quick to lay all the blame on fansubbers here. The first people to really adopt h.264 was the DVD rippers, and they definitely used it for higher quality (and larger file sizes) first, before fansub groups. (Whether we like it or not, DVD rippers are always on the front edge of new encoding technology compared with fansub groups).
I think a lot of the initial usage of h.264 in fansubs came from encoders with experience in the dvd ripping scene, and that's why many of the first groups releasing h.264 encodes did larger sizes than the xvids. All the encoders I know that are purely fansubbers and have nothing to do with dvd ripping have been against large h.264 usage from the start.

So I think we can blame the mistakes you say fansubbers made in the beginning on the blind application of dvd ripping standards to fansubbing (when the two have quite different purposes).

[Mentar]

With all due respect, but the "adoption rate" of h264 encodes has pretty much NOTHING to do with filesize or the way it was "used". NOTHING. I rarely make broad sweeping statements like these, but I'm VERY sure about that here.

In the 18 months of dual-releasing avi and mkv encodes at the same time, I can't remember a _single_ instance where someone came to the channel and asked "what's the smallest file, avi or mkv?" - the question is always "what looks better, avi or mkv?"

I'm not saying that there aren't people who are concerned about the filesize at all, of course they do exist. But they are the exceptions to the rule. Here are some hard data facts to all the theories lately:

Download numbers for "new" TV-based episodic releases are consistently around 3:1 avi:mkv, and they have been like this pretty much for the entire last year. When the releases are DVD-based rereleases of formerly used material, the ratio can get below 2:1 (here, more leechers are concerned about quality).

These 3:1 numbers don't change even when the mkv releases are "split" into a large higher-res and a small lower-res version. Here, the larger HD-version has a 2:1 advantage, however, the smaller version is generally not picked for filesize considerations, but rather because people don't have sufficient CPU power to play the higher-res version back. Again, the clearly expressed main consideration from the mkv leechers is "better quality".

So why does avi still command this clear 3:1 advantage over mkv? The reason given is almost universally "interoperability". Avi works everywhere, especially also in most standalone players. For people, this really matters. Burn it on CD/DVD, toss it into your standalone, watch on the living room couch on the TV screen. Watch it on Macs, watch it on any obscure platform, avi works. And for most people, the xvid avi quality is absolutely sufficient. And as long as there is nearly no standalone playback support for mp4 or mkv (except for hacked xboxes, and it's incomplete even there), avi will keep the lead in downloads. When this changes, the numbers will change. No sooner though.

But the most important point is "does it matter to us"? Frankly, as much as I prefer mkv over avi, I accept the reasons of the avi camp. And it's not as if I'd have sold my soul to h264/mkv-mp4. It's no religion, it's merely a _tool_ for spreading anime. It enabled me to improve the quality of my work drastically over xvid avi, and a very strong minority (25%) think so, too. THEY want me to go for "best quality achievable with the raws at hand", and I'll definitely continue to do so. And in my working process, it's a trivial issue to also provide a xvid avi version at the same time. There's nearly no extra work involved, just 40 minutes of encoding and 40 minutes of uploading. In modern days, distro issues have virtually disappeared, so fans are given a choice. Want the avi? Be my guest! Want the HQ mkv? Hey, enjoy!

Theoretically, I could add a third or even fourth release - as small as possible while retaining around xvid quality. But this would be overkill, and I predict it would be nearly ignored. I make sure that the entire anime fits on one DVD (better than having to juggle with two). Whether or not some remote fan with a really poor connection needs to wait 30 minutes longer to download a file does not justify YET ANOTHER version.

So, to wrap it up: Whoever believes that going the "as small as possible for the same quality" way would have increased the spreading of h264 encodes didn't talk to the fans very much. You guys should release normal-size avi and small-size mp4 at the same time and have a look at the download numbers. You'll be surprised.

[[darkfire]]

For the past few months I have been going around and talking to as much anime fans as I could in my city and surrounding areas. This is what I found.

"I don't have broadband...", Yes even today many people do not have broadband. I have encountered many that do not have it. I don't have it either. Why is that. Is it because its not available or they can't afford it. The latter is mostly the case. Thet just can't afford broadband. They have access to it but not in their house. The access that they do have is restricted not by bandwidth but by time. Many of these people go to a community college or use their local library's to download their anime. So they have to find they quickest way to get their show in the time alloted to them. What would those ways be. Truly its only two ways. Direct Downloads and Flash.

On the flash side, Anime-Eden and Youtube fulfill their needs. They can watch their show immediately. Two shows in one hour is acceptable for them because its faster than downloading the file. On the DDL side thats where all the confusion starts. They have no knowledge of containers or codecs. Most ddl sites have both the original file and a rmvb file. The rmvb is usually smaller averaging around 30-60. This file size allows them to get their files fast and get on their way home. Also the rmvb file quality is not bad. It isn't as good as the original but the majority of the users don't care. They just want to fill their thirst for anime.

Whats the problem? The downloads associate small file sizes to the rmvb extension. So no matter what you encode your file with if its not rmvb they won't get it. There is also another problem. They are mostly using "restricted" computers. What does that mean. No IRC and no bittorent and with the majority of ddl sites hosting rmvb a fansubbers changing codecs will do nothing significant. Its a start but their are some issues that need to be solved.

Change is in the air. We owe all to the ipod and psp. They use the mp4 container. They have been exposed to h264 because of this. But most of them don't know that is h264 they just know that it looks nice. Some ddl sites are changing from rmvb to AVC/MP4 because of in order of importance, 1 smaller size, 2 better quality.

Why is this important? Smaller sizes are easier to host, easier to distribute, are faster to download. This allows the "leecher" to watch more shows and fill their thumbdrive (they use sneakernet to move files from school/library to their house) slower. For example, one of my friends downloading Love Hina in and love Hina again in 2 days. All of the files were rmvb and varied from 30-60megs.

What we need to do. Encoder with smaller sizes. I know to the encoder quality is key. But quality at the expenses of filesize should not happen. Their needs to be a balance. Situational based encoding seems to be the best answer. Froth-bite is doing well with their encodes of 140megs its great but the majority of leechers wont download anything over 70megs. Can we do encodes like this. Yes, as thefluff pointed out NHG was 80megs. Its possible to get small sizes. We should all try to. I know I will.

Edit:

Originally Posted by Mentar

So, to wrap it up: Whoever believes that going the "as small as possible for the same quality" way would have increased the spreading of h264 encodes didn't talk to the fans very much. You guys should release normal-size avi and small-size mp4 at the same time and have a look at the download numbers. You'll be surprised.

I'm a fansubber previous leecher. For the "Majority" of leechers its all about size not quality. The "Majority" of leechers don't get on irc, don't use torrents, and don't know that Animesuki even exists. Their time restricted internet usage forces them to want the smallest size. What we encounter is the tail of the "Majority" that can get on irc or use bittorents. That tail seems small but behind the wall is a dragon. The internet is not the place to find your stats. Its the real world.

[Mentar]

Originally Posted by [darkfire]

What we need to do. Encoder with smaller sizes. I know to the encoder quality is key. But quality at the expenses of filesize should not happen. Their needs to be a balance. Situational based encoding seems to be the best answer. Froth-bite is doing well with their encodes of 140megs its great but the majority of leechers wont download anything over 70megs.

Where do you have all this stuff from? What we have here is alot of highly questionable opinion, but no trace of proof.

My friends at Froth-bite may please forgive me, but I honestly see no indication that they are doing particularly well with their _sole_ 140 meg h264/mp4 releases. I can't remember any of their releases breaking the 10.000 downloads barrier since they switched away from avi (which they broke regularly back then), while their avi-touting competition breaks 20.000 with ease, and releasing much later! Also, I want to be on record that I consider F-B releases to be hi-quality. So, by the empirical data, I can't see how it would agree with you.

Also, where the fish do you have the 70 megs limit from? Good lord. I can't even remember seeing any fansub release below 100 megs in ages which wasn't a 2-minute DVD extra.

Can we do encodes like this. Yes, as thefluff pointed out NHG was 80megs. Its possible to get small sizes. We should all try to. I know I will.

Feel free to do so. I won't.

I'm a fansubber previous leecher. For the "Majority" of leechers its all about size not quality.

Any trace of proof for THAT? I strongly doubt this claim. In my experience it's about a) speed, b) format and c) quality, in this order. Size has never had any impact on release numbers.

The "Majority" of leechers don't get on irc, don't use torrents, and don't know that Animesuki even exists. Their time restricted internet usage forces them to want the smallest size. What we encounter is the tail of the "Majority" that can get on irc or use bittorents. That tail seems small but behind the wall is a dragon. The internet is not the place to find your stats. Its the real world.

One day, when you happen to find this dragon, please don't forget to tell me where it is.

In the meantime I'll continue to concentrate on the tail

[[darkfire]]

The proof I live with them talk to them everyday. But they don't get on IRC or here so I can't show you. If they did my it would invalidate my point ^_^

[Sylf]

Originally Posted by [darkfire]

For the past few months I have been going around and talking to as much anime fans as I could in my city and surrounding areas.

Where is "my city?" That's a dangerous way to try to summarize what the "average leecher" of fansubs want.

I live in US, in a med-small sized town (population 300,000). I don't know of many people who don't have broadband in my area. And the broadband availability is quickly propagating to rural areas too. So, if I take my area as example, I'd say that I can safely assume that average people (especially people who want to watch fansubs) have their own access to broadband.

Even after reading what could have been a semi-extensive analysis of the situation, it doesn't have any weight of credibility. We're distributing our digisubs over the internet, so show me the desires of the average internet-user fansub watchers if you want to sound at all convincing.

Just for the record, when Froth-Bite made a hard switch to h.264/mp4 format a little over a year ago, we only had to assist a handful of people to get them set up to watch our releases. If the requirement to watch those files were truly cryptic, then I think we would have been more overwhelmed with questions from those leechers.

[[darkfire]]

I'm just south of chicago. Park Forest if you want details. http://en.wikipedia.org/wiki/Park_Forest,_Illinois

Originally Posted by Sylf

Just for the record, when Froth-Bite made a hard switch to h.264/mp4 format a little over a year ago, we only had to assist a handful of people to get them set up to watch our releases. If the requirement to watch those files were truly cryptic, then I think we would have been more overwhelmed with questions from those leechers

My answer

Originally Posted by darkfire

The "Majority" of leechers don't get on irc, don't use torrents, and don't know that Animesuki even exists. Their time restricted internet usage forces them to want the smallest size.

I was just showing the other side. Yah I have time restrictions on my internet usage but I know ways around it. I really don't care how its encoded. I like my stuff h.264 and HD res. Thats just my personal preferences, but many around me don't agree. For example my brother only gets avi. Cause he can convert them to his psp easily. I for one dont see why he can't convert mkvs, or softsubs, but then I am an encoder so my view is biased.

[Sylf]

I'm sure that's the case among the sample you've seen around your neighborhood. And I'm sure there are other neighborhood like that around the world. But do realize that it's just a very small part of world you've exposed yourself to, and it's dangerous to conclude that as "majority." I'm not doubting the result of your observation.

[Skane]

Just some personal input. When it comes to fansubs, I look out for the following in descending priority,

..) Translation Accuracy

If one group had 233mb encodes with excellent translations, as opposed to another group with 175mb encodes with questionable translations, I would go with the former in a heartbeat. This remains my "de facto" deciding factor in choosing which encode to download. Filesize and file-format be damned.

..) Softsubbed

Under the circumstance that both encodes are of equal translation quality, the next thing I look out for is softsubs. This mainly means the .MKV container.

..) Video Quality

This one is self-explanatory. Call me old-fashioned, but I simply cannot phantom why some people would want to watch anime on YouTube when most of them have the means to download a better-quality version. I know a lot of friends who have better broadband than me, and they still prefer to watch from YouTube.

..) Filesize

Ah... we finally reach the current point of the discussion. Only after the above 3 points have been addressed, do I actively consider this point. This is one of the reasons why I really love [Froth-Bite]'s releases. The only thing missing from them are softsubs( they are currently the closest to perfection in my personal opinion( for a fansub group that relies on TV-RAWs)).

Cheers.

[Starks]

People will watch fansubs regardless of quality.

[DryFire]

Originally Posted by Skane

Just some personal input. When it comes to fansubs, I look out for the following in descending priority,

..) Softsubbed

....

...) Video Quality

Are you sure that's in the right order? Unless you planned to take the softsubs and watch the raw. Thankfully, most people who release softsubs care about video quality.

As a side note, softsubs do not require mkv; mp4 supports softsubs but in a format I've yet to see used (no point in bringing up ogm :/ )

[Devastator]

If you're referring to TTXT, I took a quick look at some sample scripts and I must say, ASS/SSA looks much simpler to use in my opinion. Considering all the different tags (FontTable, FontTableEntry, Style, TextSample, Karaoke, KaraokeRange, etc.), I for one wouldn't want to be bothered to use it.

Otherwise if you meant SRT, I think the given response is probably in regards to styling. As far as I know, fonts can't be included in MP4 files, which would have rendered SSA and ASS pretty much useless anyways, if they were options.

XML, IDX and SUB, I wish I could say something about, but I know little to nothing about them aside from the fact that TTXT appears to be essentially an XML file.

[DryFire]

I meant softsubs does not imply mean mkv only.

[Devastator]

Don't worry, I understood that part. (:

I was referring to this:

Originally Posted by DryFire

mp4 supports softsubs but in a format I've yet to see used

[Quarkboy]

Originally Posted by Devastator

Don't worry, I understood that part. (:

I was referring to this:

ttxt is NOT meant to be something that is easily human editable directly, like ssa is.

gpac has defined an xml based wrapper for it, which is something that's easily PARSABLE by a computer (which, by the way, ssa is not easily parsable... Just ask any of the developers of vsfilter). So dealing with that sub format really requires using some software that can output to the file format, which, at the moment, does not exist. The only thing one can do so far is to to import .srt format subs with mp4box and to export the .ttxt format. But since .srt contains very little information, it's not particularly useful.

Actually, I experimented a bit with mp4 softsubs using that format for my Yes! Pretty Cure 5 subs (since they were sort of designed to be fast), and I got a lot of it working. But, oddly enough, the only program that would display the subs properly was vlc! Haali's media splitter's support for the subtitles was not working right or I was doing something wrong.

(How amusing that vlc is the only program that displays the subs right...)

[Kurth]

About the filesize and preference of the leechers.

Well I was looking at the number of the complete downloads of that new anime Juusoukikou Dancougar Nova that have one version encoded with Xvid and another one encoded with H264 but those two files have a good difference on filesize.



Xvid version 233 MB resolution 640x360 number of complete downloads 3.464
H264 version 170 MB resolution 640x360 number of complete downloads 12.933

Looking at this I really think people prefer to download small files even if those files are encoded with the H264 codec, they want to watch it as fast as they can soo downloading a small file is better.
It is not like they want the best quality ever they just want to waste less time to download because the 170 MB file is enough to watch the anime, they dont need the 233 MB file. Only people that hate H264 , have a very old computer and want an easy way to convert to another format downloaded the AVI version.

[Quarkboy]

Originally Posted by Kurth

About the filesize and preference of the leechers.

Well I was looking at the number of the complete downloads of that new anime Juusoukikou Dancougar Nova that have one version encoded with Xvid and another one encoded with H264 but those two files have a good difference on filesize.

...

Although I'd like to agree with you, your evidence is flawed. Looking at animesuki's records, the mp4 version was released 1 day before the xvid version.

Such a comparison is only really valid if the files are released simultaneously.

[Devastator]

Originally Posted by Quarkboy

ttxt is NOT meant to be something that is easily human editable directly, like ssa is.

Well, I figured that much. Looking over the sample scripts, I think it probably wouldn't take too long to be able to gather ideas on how to create the output module, depending on how the input is generated.

I'm thinking the nicest it could get would probably be a GUI interface that displays the information as a tree, from which branches can be inserted (either with some input box asking for the necessary parameters or perhaps a button that just adds a line with empty parameters) / deleted / duplicated and trees can be collapsed / expanded. This way, if a user removes a branch or tree, the user can see how much information is lost.

I must applaud for your effort in using the format though, it certainly isn't something that I would be entirely eager to do.

====

On the subject of the quality meter, it can vary depending on the situation. If there's a group working on said show that I respect, chances are that I'll follow their release over any of the competitors.

I regard video quality as being more important than softsubs. Between translation and video quality, I can't really say. I don't know the language as well as I would like, so I usually don't mind too much as long as the spelling and grammar is mostly correct and the translation retains the information relevant to the story. Since I care about video quality though, I try to obtain the h.264 version when possible.

Next would probably be softsubs. I'm not entirely picky whether a group chooses to use softsubs or not, unless there's something about the translation that gets at me (then I get rather frustrated about not being able to change it).

Lastly, file size. In my case, file size doesn't pose a problem for me since I have DSL and no bandwidth cap. The only time that file size would probably get at me is if the series if filling up a good portion of my HDD space. Nothing DVDs can't help with though for me (considering how much I cut down on watching anime over the past year).

[Skane]

Originally Posted by DryFire

Are you sure that's in the right order? Unless you planned to take the softsubs and watch the raw. Thankfully, most people who release softsubs care about video quality.

As a side note, softsubs do not require mkv; mp4 supports softsubs but in a format I've yet to see used (no point in bringing up ogm :/ )

I didn't say it was only .MKV, I said mainly .MKV because that is the most common format used by fansubbers for softsubs. I have played other formats as well, such as .OGG and its' ilk.

I am quite comfortable in adopting new formats so long as the codecs( and whatever else is needed) are easily searchable and do not require me to jump through loops of fire in installing them.

Besides, I am not representing any group. I did say that it was just my own personal preferences.

Cheers.