Alternative matroska splitter for directshow

[gabest]

Those who are interesting in subtitling, could you please test the new subtitle source filter? With it now you can put ssa/ass into matroska! Don't forget to update the matroska muxer/splitter and vsfilter, mpc for playback.

[Assault]

Great work Gabest!!!

The strange square that could be seen at the end of the first subtitle line when two lines were displayed is gone. :D :D :D

Assault

[Liisachan]

Originally posted by gabest
With it now you can put ssa/ass into matroska!

w00t! metacool
You are beyond greatness!!!!!! I dont know what to say. I ll test it today

[bill_baroud]

i'm testing with Gabest on IRC and it have some problems with crazy effect ssa and my 120fps clip ....

ah ?

[23:04] <Gabest> ahhhhhhhhhhhhhhhhhhhhhh
[23:05] <Gabest> I used the wide char version to print an 8bit string
[23:05] <Gabest> style names are only stored as ansi
[23:06] <Gabest> bill_baroud: I debugged the unicode build and released the non-unicode "for the masses" :P


looks like we need a new release :)

[gabest]

Hm, just wanted to write the same thing :)

Subtitle Source 1.0.0.1 released.

[ssjkakaroto]

thx a lot for these updates gabest, now the square really is gone :D

OT: @bill_baroud: what is the deal with those 120fps videos? is there really the need for such frame rate?

tia

[Animaniac]

Is the Morgan Stream Switcher currently viable for stream switching ? Last time I tried it, it broke all DShow playback... Gabest, do you plan on implementing stream switching in the Matroska Splitter? Since, that would be great!

[gabest]

No, there won't be any stream switching in the splitter, that way streams have to be identical in encoding and every parameter.

I'll check mmswitcher. edit: Seems to work as long as you have audio streams with the same channelnumber/frequency/bits.

[Tueurne]

Maybe it's not the priority, I don't know if it's hard to implement but could it be possible to have the chapters support in decoding Gabest and the name of audio language.
I think it cames from decoding like blacksun told me.
(I use last VDM CVS to mux)

[bill_baroud]

@ssjkakaroto : it's the way japanese crazy encoders handle Hybrid clip ... do the math, 24x5 = 30x4 = 120 ... there is 3/4 dummy frames in the avi between each "true" frame. But it create big overhead (~10mb for 25min) and it consume more cpu power (at least for me)

Matroska is based on timestamp and doesn't need those ugly tricks :D

[Tueurne]

Originally posted by Tueurne
Maybe it's not the priority, I don't know if it's hard to implement but could it be possible to have the chapters support in decoding Gabest and the name of audio language.
I think it cames from decoding like blacksun told me.
(I use last VDM CVS to mux)

Sorry I find my answer.
It doesn't come from decoding...

[Kyo]

gabest great work with the ssa subs, the filters work like charm

[report]
with the new vsfilter(2.26) the problems with subs on cd are still present :(

Thanks for all the work
;)

[Liisachan]

Originally posted by gabest
Those who are interesting in subtitling, could you please test the new subtitle source filter? With it now you can put ssa/ass into matroska! Don't forget to update the matroska muxer/splitter and vsfilter, mpc for playback.

A test Matroska clip with 2 (Unicode) SSAs embedded has been muxed successfully. This is a multi-language SSA softsub, in English and in Japanese.
It plays perfectly with Media Player Classic!
Check this Graph (http://ld-anime.faireal.net/graph.png) and the result (http://ld-anime.faireal.net/mpc.jpg).

More information/pics about the test result. (http://matroska.tripod.co.jp/ssa.html)

My compliments to you, Gebest!:) :) :) :)

[defeatist]

This is fantastic! Please keep up the excellent work, Matroska Team.
My sincerest thanks go out to you especially, gabest, for everything you continue to do for this truly edifying project. We are all indebted to you.

I am certain Matroska's userbase will expand greatly once awareness is increased. All that's preventing many encoders I know from using it is the lack of chapter support. Bless us with that, and MKV can replace OGM entirely (IMHO). In fact, Matroska should claim far more users that OGM has. That is my prediction, at least. Please consider it.

Thank you and good luck!

[ChristianHJW]

What can i add here, Gabest made a lot of people really happy i guess :) !

I personally wonder when people will start capturing with the muxer filter into MKV, this should work out of the box with any WDM capture driver already.

I made a first ( crappy ) example graph here :

http://matroska.free.fr/capture/capture.JPG


Oh, and BTW, for completeness :

Mosu's latest mkvtoolnix 0.5.0 allows muxing of SSA subs into MKV from CLI also ;) ....

[Liisachan]

Originally posted by ChristianHJW
Mosu's latest mkvtoolnix 0.5.0 allows muxing of SSA subs into MKV from CLI also ;) ....

Actually I first tried MKVmerge 0.5.0 but it freezes after it says "Using SSA/ASS subtitle reader", or it just says "Error: File eng.utf16.ssa has unknown type." Altho it can mux SRT subs perfectly, Maybe my cmd was wrong... so this time I used MKVmerge 0.5.0 only for writing the language tag (jpn)...

BTW: even if there is an language tag, MPC and VobSub shows all sub tracks as "embeded" without showing the language name. Hope this will be fixed soon ^^;

[gabest]

Originally posted by ChristianHJW
I personally wonder when people will start capturing with the muxer filter into MKV, this should work out of the box with any WDM capture driver already.Try it with MPC, it can capture to matroska ... at least with the btwincap drivers :)

Liisachan: I'm not sure, but maybe it only reads utf-8 text files.

[Liisachan]

Originally posted by gabest
Liisachan: I'm not sure, but maybe it only reads utf-8 text files.
Possibly... to begin with --sub-charset doesnt recognize UTF-16, if i m not wrong.
I had already tried more than once SSA in UTF-8, with --sub-charset n:UTF-8 switches, or in the def CodePage (Shift_Jis for me) without the --sub-charset switch. But it didnt work so far. I'll retry later, as probably i typed wrong cmdlines. How to use the TID params in mkvmerge 0.5.0 is a bit complicated...:confused:

Meanwhile, here s a little bit more test results about your alt filters.
On my windows 2000 SP3, this multi-SSA softsub MKV file (DivX5+Vorbis) plays fine in: Media Player Classic 6.4.5.6
Windows Media Player 9
Windows Media Player 6.4
Zoom Player 3.10.1
Sasami 2K #757
Light Alloy 2.4
It doesn't play in: TCMP 4.0 RC3 Matroska Edition
BS Player 0.86 #499
Micatoge XPlayer v2.2.4
XPlayer plays video only, but no audio/subs.

[Bluedan]

I don't get it:

bsplayer v086.500 and TCMP close immediately after opening mkv file without warning while others (Zoom, wmp) don't.

I installed latest filters from gabest (16-6-2003), already uninstalled the mkxds.dll (0.4.3), corevorbis.dll from 10-06-2003, matroskaprop(erties).dll, cdl-plugin for use of a mkv muxed with latest VDM build 1639 containing XVID and 1 oggvorbis audio. Every filter properly registered.
I also updated oggvorbis dll's, but that should not cause any trouble.
Nobody complained about it so far, so I guess I'm the one to blame.
What am I doing wrong? What makes TCMP and bsplayer commonly different?

[ChristianHJW]

What audio has your mkv file ?

[MarkCoolio]

This is what made me play mkv files with TCMP:

- Video: Disable Built-In Subtitle options within TCMP (2 checkboxes)
- Audio: Disable Dynamic Equalizer

Hope this helps.

[ChristianHJW]

Originally posted by MarkCoolio This is what made me play mkv files with TCMP:
- Video: Disable Built-In Subtitle options within TCMP (2 checkboxes)


Now as you mention it, i guess BSplayer has its own subs renderer also, like TCMP, so i guess this is the reason for the crashes ....

[Bluedan]

Originally posted by ChristianHJW
Now as you mention it, i guess BSplayer has its own subs renderer also, like TCMP, so i guess this is the reason for the crashes ....

Hmm, no subtitles in use with that mkv. Audio is 2 channel ogg vorbis q=0.200
I wonder about that special matroska support in TCMP (CDL-plugin), if it causes any trouble.

[ChristianHJW]

Originally posted by Bluedan I wonder about that special matroska support in TCMP (CDL-plugin), if it causes any trouble. .. as a simple test ( maybe you have an old version of the CDL ? ), simply remove it from the \TCMP\plugins directory ( move it to \TCMP\ root temporarily ) and restart the player ....

[Liisachan]

Originally posted by Liisachan
Actually I first tried MKVmerge 0.5.0 but it freezes after it says "Using SSA/ASS subtitle reader"

I can't solve this prob after all.
mkvmerge -o out.mkv in.avi in.ogg en.srt does work,
while mkvmerge -o out.mkv in.avi in.ogg en.ssa does not, even in a simplest case such as
mkvmerge -o out.mkv in.avi in.ogg --sub-charset 0:ASCII --language 0:eng en.ssa

Here's the output:

C:\mkvtool>mkvmerge -o out.mkv in.avi in.ogg en.ssa
Using AVI demultiplexer for in.avi. Opening file. This may take some time depend
ing on the file's size.
+-> Using video output module for video track ID 0.
Using OGG/OGM demultiplexer for in.ogg.
OGG/OGM demultiplexer (in.ogg): using Vorbis output module for stream 23438.
Using SSA/ASS subtitle reader for en.ssa.
+-> Using text subtitle output module for subtitles.
Opened 'out.mkv' for writing. <--it stops here :(
^C

--i m guessing 0.5.0 can read SSA but cannot write S_TEXT/SSA on my win 2k:confused: Does this have anything to do with these alt filters? do i need the official MKxDS.dll to run mkvmerge or...?

[ssjkakaroto]

Liisachan i've already reported this behaviour to Mosu, i think in the next version this will be fixed :)

[Bluedan]

@christianHJW

that does not solve it, besides it won't explain the exact same behaviour in bsplayer...

EDIT:
this is what graphedit is showing: graph (http://www.shoobiedoobiedoo.de/pub/graph_mkv.JPG)

BTW, why is ogg_submuxer always kicking in?

view chunk. (http://www.shoobiedoobiedoo.de/pub/rossi_chunk.mkv)

[Liisachan]

@ssjkakaroto: I see, thx^^

BTW: does this "soft ssa" in Matroska alread work on Win98 too?

[Bluedan]

more preceisely: the mkv file refuses to play in tcmp and bsplayer when ogg audio is in use. Tested both with wav and mp3 and they played perfectly.

Is Core Vorbis to blame? Toff?
Anyone confirm this?

[[Toff]]

I think there can be a problem with CoreVorbis decoding to 32 bits and the visualization filter or the DRC filter. Try to force 16 bits decoding in CoreVorbis or to "turn off" the filters.

[Animaniac]

Originally posted by Bluedan
@christianHJW

that does not solve it, besides it won't explain the exact same behaviour in bsplayer...

EDIT:
this is what graphedit is showing: graph (http://www.shoobiedoobiedoo.de/pub/graph_mkv.JPG)

BTW, why is ogg_submuxer always kicking in?

view chunk. (http://www.shoobiedoobiedoo.de/pub/rossi_chunk.mkv)

It's meant to always kick in. There no need for SubTitDS, since VSFilter handles OGM perfectly now. I've uninstalled SubTitDS.

[Liisachan]

Tried the newest filters, MatroskaSplitter 1.0.1.2 & VSFilter 2.27.
Now the subtitle menu for a multi-subbed clip is neat :D like these pics (http://matroska.tripod.co.jp/splitter1012.html) . I'm so happy to see this :) :).

However, apparently real-time effects in soft-ssa such as Fade-in/out or Karaoke don't work with MPC's intrinsic subtitle renderer, while they work fine if you use DirectVobSub after unchecking "Pre-buffer subpictures."

Win2K + Dx9 + MPC6.4.5.7, useing VMR9 (renderless)

[gabest]

In mpc you can find the same option but in a bit different form. If you set the number of subpictures to prebuffer to zero, it will have the same meaning as turning it off. But don't expect good results in fullscreen at some high texture resolution :)

[Liisachan]

thanks for the tips, it works *^_^* what's more, now MPC works best for me for Karaoke etc.

Like you said, I have to lower "maximum texture resolution" to work un-buffered subs smoothly in my old PC (P3 850),
but the result is decent--less cpu-intensive than DirectVobSub (with or without "Doubled resolution"). I really like this feature where you can choose "resolution" so the result will be best with your cpu power, video environments etc.

I also found that subs by MPC in fullscreen mode is surprisingly beautiful if "maximum texture resolution" is high...unlike other players, MPC's subs are not "drawn on video then resized with video" but "video resized, then subs drawn on it" (right? ...that is my observation) and so subs are high-resolution.

[filewalker]

I'm also happy that DVobSub shows the language names. THANKS gabest!:)

I generally use latest VDMod for muxing MKV.

After some tests with the new VSFilter 2.27 I found a little bug, I think.

I added a .srt subtitle in VDMod into the streamlist, then I pushed the "Comments" button for the .srt file.
Then I chose "language" from the list and used the appearing language selection.
If I choose from this predifined language selection, DVobSub displays the selected languages correctly.:)

But If I don't use this predefined language selection and write manually e.g "french (normal caption)" instead, in the "value"-line, DVobSub displays just "English" for each subtitle although I never typed into "English"
(This method with typing into the "value"-line works fine for OGM in VDMod.)

Is it a bug or is it not allowed anymore for MKV to type in manually the comments?

Cu filewalker

[gabest]

Liisachan: mpc renders the subtitles in the actual window resolution or the max texture size, whichever is smaller, while dvobsub can only use the video's resolution. This was one of the reasons I made my own player.

filewalker: When the language element of the track is empty dvobsub defaults to "English", I'm going to ask cyrius how vdubmod saves mkv when you don't pick a predefined language.

[Liisachan]

Originally posted by filewalker
Is it a bug or is it not allowed anymore for MKV to type in manually the comments?
This is my question too. My friends from Brazil will be unhappy because there is no tag for Brazilian in the ISO639-2 code table generated by mkvmerge. Sometimes users should want like "Portuguese (Brazilian)" "English (by Foo)" "English (by Bar)" or "w00t (l33t $pe@k)" or "Klingon" or anything...
This might be a touchy subject but all those are possible and easy with OGM, especially if you edit OMX files manually

[filewalker]

Originally posted by gabest
filewalker: When the langauge element of the track is empty dvobsub defualts to "English", I'm going to ask cyrius how vdubmod saves mkv when you don't pick a predefined language.
For closer explanation:
In the first line I choose "language" from the list.
After that the language element shows "User defined".
Then I type my desired comment into the "value"-line beside the "language element" line.

Thanks gabest, that you take up this problem or talk with cyrius about it.:)

Cu filewalker

[[Toff]]

If you don't provide a language code for a track, that mean that this track will be using the default langage which is english. So if you give a name to the track but no language code the filter will display english.
Also i don't think that the track name is displayed anywhere yet.

Now there is maybe another problem if there is no iso code for unknown langage, i haven't checked it yet.

[filewalker]

Originally posted by [Toff]
Also i don't think that the track name is displayed anywhere yet.
You can test yourself, the track name is shown well, if you mux the "typed in" comments into OGM container.
...it only doesn't work in MKV (muxed with VDMod).

To type in the comments is very useful:
Just imagine this example:
If you have 2 subtitles in French.The first one of them is "normal caption" and the other one is "forced caption".
ATM you can't distinguish between them because both subtitles are only displayed as "French"

In OGM I used to write for the first sub "french(normal caption)" and for the second "french (forced caption)".

Cu filewalker :)

[Atamido]

Originally posted by Liisachan
My friends from Brazil will be unhappy because there is no tag for Brazilian in the ISO639-2 code table generated by mkvmerge. Sometimes users should want like "Portuguese (Brazilian)" "English (by Foo)" "English (by Bar)" or "w00t (l33t $pe@k)" or "Klingon" or anything... Thats because all of those are dialects, not languages. In the first case, you would pick the language "Portuguese", and then label in Brazilian. The same is true with "w00t", which would be a dialect of probably English, and then label it (l33t $pe@k). I don't think that there is anything that you could pick for Klingon as its not a real language...

Anyway, hopefully the filter will show the description soon.

[Liisachan]

Originally posted by Pamel
Anyway, hopefully the filter will show the description soon.

Yep. If the filter shows the description, there will be no problem.

I'm using OGM's language tag not only to tag the language but to tag description, because DVobSub will show it anyway like this (http://ld-anime.faireal.net/tmp/ogm5menu.png), and filewalker means the same thing, I think.

You said Brazilian is just a dialect, well, you could say so.
But realistically, for subtitle labels, it would be inconvenient if you could not tag "Simplified Chinese" and "Tranditional Chinese" or "Mandarin" and "Cantonese," even tho you could say "all those are dialects of chinese."

On the other hand, "Norwegian" "Nynorsk" and "Bokmål" has each own code, which is a bit strange. Altho, Gabest is not the one who decided ISO 639-2 Language Codes (http://www.loc.gov/standards/iso639-2/langcodes.html). Even tho Klingon was rejected, there will be much more languages in Unicode Tables than in ISO639-2, and unlike OGM, which uses old and inconvenient MS Code Pages, Matroska is Unicode-based.

Check Unicode Roadmap (http://www.unicode.org/roadmaps/), Proposed New Scripts (http://www.unicode.org/pending/pending.html) and Proposed New Characters (http://www.unicode.org/alloc/Pipeline.html) , which includes Proposal to encode Tengwar in Plane 1 of ISO/IEC 10646-2 (http://std.dkuug.dk/jtc1/sc2/wg2/docs/n1641/n1641.htm)--I'm sure some Tolkien fan subbers would love to use these code points to sub Elen Sila Lumen... or something like that :)

[celtic_druid]

@Pamel, not a real language?
http://www.amazon.com/exec/obidos/tg/detail/-/067174559X/102-2714788-6235334?vi=glance

*Note: I don't actually own a copy or any other Klingon literature.

[Atamido]

That book lists less then 200 words in Klingon. If it truly is the "comprehensive sourcebood for Klinglon language and syntax", then I don't see the language code for Klingon seeing much use.

@Liisacha: Good point. I didn't notice that there weren't seperate language codes for Mandarin and Cantonese. We really need to get the track descriptions working.

[HomiE FR]

@gabest: First, thanks for your wonderful DShow filters. Then, I have a little problem here with latest MatroskaSplitter/MatroskaMuxer (don't know which one, maybe both): I've done an ASS subtitle file with 3 embedded custom fonts (using Sub Station Alpha) which works perfectly when it's put in the same directory as the audio/video file (a MKV file here). VSFilter 2.27 uses the embedded fonts (since I uninstalled the fonts and tried to play the file+subtitles on another computer without these custom fonts installed).

But then I tried to mux this ASS script (first the file was ANSI, then I tried with an Unicode one, but it didn't change anything, which is normal since there were only "standard" characters in this subtitle file) within a Matroska file, with an XviD video file (from an AVI) and an Ogg Vorbis audio file. I used the latest DShow filters on guliverkli with GraphEdit and there wasn't any problems.

But then, when I read the Matroska file with the ASS script in it, it won't read my embedded fonts, it just put the standard Arial. So I guess that either SubtitleSource (ASS) or/and MatroskaMuxer or/and MatroskaSplitter don't check the [Fonts] events inside the ASS file, because in both cases I read the subtitles with VSFilter 2.27 (Unicode) so if it works with the ASS script outside the Matroska file it should work with the script in it.

I'd really like to see those embedded fonts working even when included inside a Matroska file. :) I don't know if it's hard to implement (or if I did anything wrong) but I think it's worth it. I hope I didn't miss anything obvious.

My versions:
MatroskaSplitter 1.0.1.2
MatroskaMuxer 1.0.0.6
VSFilter 2.27 (Unicode)

Thanks in advance.

[gabest]

Yea... this isn't working, and the problem is the subtitle source filter. It first reads in any subtitle file it knows and converts its data into its own inner format, then generates the new ssa file header for matroska from this information. Unfortunatelly embeded fonts aren't stored and written, because immidiatelly when they are first read they get converted to a ttf file in the default temp folder of windows and loaded into the system for usage. In the future I'm going to store the font data and also rewrite it into the new file header, but currently it is just not working that way.

[HomiE FR]

@gabest: Thanks for the explanation. I know the guilty one now. :) No problem, I'll wait for a new SubtitleSource version. However, thinking of the way you're talking about it, it seems it's not as easy as I first imagined. Too bad, but be sure I'll check for your updates.

[celtic_druid]

Well I believe that they somehow managed to translate the bible, hamlet and some other shakespear works. So I would think you could do Klingon for pretty much anything if you really wanted to.

[Liisachan]

Still testing the guliverkli filters, they are fantastic!

Upped more screenshots (http://mypage.naver.co.jp/freenet/matroska/ssa/) for cool Unicode subtitles :p

However, I think MatroskaSplitter 1.0.1.2 has this problem:

(1) If the path to the MKV file includes a double-byte character,
any players except MPC can not open the file with MatroskaSplitter.ax,
getting "unknown error."

(2) If MKxDS.dll is used instead, there is no problem. So I believe this is a problem related to MatroskaSplitter 1.0.1.2

Example:
Assume you had a file c:\foo\##\bar.mkv where ## = a double-byte character (such as a Japanese letter)
- If you used MatroskaSplitter for MKV, a player like WMP9 WMP6.4 TCMP would fail to open bar.mkv getting "unknown error"
- If you unregistered MatroskaSplitter and re-registered MKxDS.DLL, the same player now could open bar.mkv properly

I may be wrong, but this is what I think I m experiencing now:(

[gabest]

It is because I release the non-unicode build, but if there is a need I can do the same as with mpc, vobsub, ... and make a 2k/xp version too.