Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[x_xy_y2]

small issue
http://i.imgur.com/TaTjJ.jpg

the line as it appears in the .ass file is:
Dialogue: 0,0:16:04.74,0:16:08.58,Default,,0000,0000,0000,,U–Um, that's what we thought,

it doesnt seem to occur all the time, ive seen it happen a couple times already but i am not sure what the trigger is

it also doesnt seem to handle negative or out of frame positions for a \move very well/at all

for the first issue you show: I failed to reproduce it.
for the \move with out of range positions problem:  I don't get it. Any related script? What's wrong with the result (in comparison to official VSFilter 2.39)?

[taulin]

small issue
http://i.imgur.com/TaTjJ.jpg

the line as it appears in the .ass file is:
Dialogue: 0,0:16:04.74,0:16:08.58,Default,,0000,0000,0000,,U–Um, that's what we thought,

it doesnt seem to occur all the time, ive seen it happen a couple times already but i am not sure what the trigger is

it also doesnt seem to handle negative or out of frame positions for a \move very well/at all

for the first issue you show: I failed to reproduce it.
for the \move with out of range positions problem:  I don't get it. Any related script? What's wrong with the result (in comparison to official VSFilter 2.39)?

i am not sure how to reproduce the first issue myself, i have only witnessed it happen sometimes

for the second issue, here is a script to try: http://www.mediafire.com/?ncm3asy1dwwc82s
the latest threaded gommorah handles it correctly, the r3644 build of vsfilter i have crashes due to the \q2 tag and your vsfilter handles the \q2 tag fine but does not move the text

[x_xy_y2]

small issue
http://i.imgur.com/TaTjJ.jpg

the line as it appears in the .ass file is:
Dialogue: 0,0:16:04.74,0:16:08.58,Default,,0000,0000,0000,,U–Um, that's what we thought,

it doesnt seem to occur all the time, ive seen it happen a couple times already but i am not sure what the trigger is

it also doesnt seem to handle negative or out of frame positions for a \move very well/at all

for the first issue you show: I failed to reproduce it.
for the \move with out of range positions problem:  I don't get it. Any related script? What's wrong with the result (in comparison to official VSFilter 2.39)?

i am not sure how to reproduce the first issue myself, i have only witnessed it happen sometimes

for the second issue, here is a script to try: http://www.mediafire.com/?ncm3asy1dwwc82s
the latest threaded gommorah handles it correctly, the r3644 build of vsfilter i have crashes due to the \q2 tag and your vsfilter handles the \q2 tag fine but does not move the text

I cannot access mediafire.com, maybe you can open an issue here http://code.google.com/p/xy-vsfilter/issues/list and if the script file size don't exceed 10M then you can attach it with the issue.
By the way, I start this work based on guliverkli2 project ( https://guliverkli2.svn.sourceforge.net/svnroot/guliverkli2 ) rev.114 , and don't know which rev of mpc-hc project it corresponding to exactly. It's damn difficult to merge the current head rev of official vsfilter into my code, or merge mine into theirs. So it may contain bugs that inheritent from VSFilter 2.39 and fixed later in VSFilter 2.40. Some of such bugs would be difficult for me to track and fix.

[cyberbeing]

for the second issue, here is a script to try: http://www.mediafire.com/?ncm3asy1dwwc82s
...
your vsfilter handles the \q2 tag fine but does not move the text

I can reproduce this. In VSFilter 2.39 it works correctly during playback (DirectShow interface) and using Aegisub (CSRI interface). In XY-VSFilter it is broken during playback (DirectShow interface), but for some odd reason works in Aegisub (CSRI interface).

x_xy_y2, here is the script from taulin's link:

Code: [Select]

[Script Info]
; Script generated by Aegisub 2.1.8
; http://www.aegisub.org/
Title: New subtitles
ScriptType: v4.00+
WrapStyle: 0
PlayResX: 384
PlayResY: 288
ScaledBorderAndShadow: yes
Video Aspect Ratio: 0
Video Zoom: 6
Video Position: 0
Last Style Storage: Default
Collisions: Normal

[V4+ Styles]
Format: Name, Fontname, Fontsize, PrimaryColour, SecondaryColour, OutlineColour, BackColour, Bold, Italic, Underline, StrikeOut, ScaleX, ScaleY, Spacing, Angle, BorderStyle, Outline, Shadow, Alignment, MarginL, MarginR, MarginV, Encoding
Style: Default,Arial,20,&H00FFFFFF,&H000000FF,&H00000000,&H00000000,0,0,0,0,100,100,0,0,1,2,2,2,30,30,10,1

[Events]
Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text
Dialogue: 0,0:00:00.00,0:00:20.00,Default,,0000,0000,0000,,{\move(1050,286,-550,286)}{\q2} After I'm born I finally realize, I exist just to imitate humans.  A Vocaloid fated to sing forever.  Even if the song already existed, a programmed toy accepts it just fine.  Gnawing on a leek, looking up at the sky, shedding tears, noticing that even all that is fading.  Even songs depend on character.  With no reliable source for foundation, the place I came from already destroyed.  When everyone forgets me, I'll have no heart or anything left.  I can see the end of the world for a Vocaloid.
Dialogue: 0,0:00:05.00,0:00:07.00,Default,,0000,0000,0000,,

[cyberbeing]

small issue
http://i.imgur.com/TaTjJ.jpg

I just now reproduced an issue similar to this. I suspect the common link is the use of commas , . It would seem that they are randomly being treated as code (?) instead of part of the script, which somehow causes a loop and produces duplicates? x_xy_y2, maybe you should add special handling for commas, since something in your code must be hitting a snag on them. The question is, could this be a caching issue, or does it have something to do with the new 'Speed up when loading big ass file (with tens of thousands of line)' script parser you introduced?

Scratch that idea, as it just happened again later in the script without any comma in the line. Hmm... I guess I'll try disabling the MRU cache to rule that out. Nope, still happens with the MRU cache disabled.

The good news, is it seems like it may be reproducible (happened with the same line twice in a row), but only if I load the script from the very beginning and wait until playback reaches the problem area. If I seek close to the problem area it doesn't happen. Is it the same for you taulin?

[taulin]

small issue
http://i.imgur.com/TaTjJ.jpg

I just now reproduced an issue similar to this. I suspect the common link is the use of commas , . It would seem that they are randomly being treated as code (?) instead of part of the script, which somehow causes a loop and produces duplicates? x_xy_y2, maybe you should add special handling for commas, since something in your code must be hitting a snag on them. The question is, could this be a caching issue, or does it have something to do with the new 'Speed up when loading big ass file (with tens of thousands of line)' script parser you introduced?

Scratch that idea, as it just happened again later in the script without any comma in the line. Hmm... I guess I'll try disabling the MRU cache to rule that out. Nope, still happens with the MRU cache disabled.

The good news, is it seems like it may be reproducible (happened with the same line twice in a row), but only if I load the script from the very beginning and wait until playback reaches the problem area. If I seek close to the problem area it doesn't happen. Is it the same for you taulin?

i'm having a hard time getting it to reproduce the same line but the problem earlier was that i was seeking near the line which rendered incorrectly, so it wouldn't reproduce, but it seems to me that it will turn up sometime after about 10 minutes of continuous playback
well it happened just past the 10 minute mark in the video, but i had skipped the prelude and OP, so it was after about 6.5 minutes of playback, so that might not be connected at all
http://i.imgur.com/P6WfS.png
http://i.imgur.com/50qy0.png
http://i.imgur.com/rKM0N.png

it's worth noting that the duplicate is replacing a portion of the line, like in the first one i posted and your example without a comma, the spaces have been replaced by the portion, "U-um," in my case, "Well," and "Was" in both your cases, in my two new images a single word has been replaced, "to" replaced with "No..." in the first and "at" replaced with "Ricchan!" in the second, "is" with "three" in the third

[x_xy_y2]

small issue
http://i.imgur.com/TaTjJ.jpg

I just now reproduced an issue similar to this. I suspect the common link is the use of commas , . It would seem that they are randomly being treated as code (?) instead of part of the script, which somehow causes a loop and produces duplicates? x_xy_y2, maybe you should add special handling for commas, since something in your code must be hitting a snag on them. The question is, could this be a caching issue, or does it have something to do with the new 'Speed up when loading big ass file (with tens of thousands of line)' script parser you introduced?

Scratch that idea, as it just happened again later in the script without any comma in the line. Hmm... I guess I'll try disabling the MRU cache to rule that out. Nope, still happens with the MRU cache disabled.

The good news, is it seems like it may be reproducible (happened with the same line twice in a row), but only if I load the script from the very beginning and wait until playback reaches the problem area. If I seek close to the problem area it doesn't happen. Is it the same for you taulin?

i'm having a hard time getting it to reproduce the same line but the problem earlier was that i was seeking near the line which rendered incorrectly, so it wouldn't reproduce, but it seems to me that it will turn up sometime after about 10 minutes of continuous playback
well it happened just past the 10 minute mark in the video, but i had skipped the prelude and OP, so it was after about 6.5 minutes of playback, so that might not be connected at all
http://i.imgur.com/P6WfS.png
http://i.imgur.com/50qy0.png
http://i.imgur.com/rKM0N.png

it's worth noting that the duplicate is replacing a portion of the line, like in the first one i posted and your example without a comma, the spaces have been replaced by the portion, "U-um," in my case, "Well," and "Was" in both your cases, in my two new images a single word has been replaced, "to" replaced with "No..." in the first and "at" replaced with "Ricchan!" in the second, "is" with "three" in the third

Just read the readme file from http://xy-vsfilter.googlecode.com/issues/attachment?aid=90000001&name=duplicate0.png&token=033d80ce84983d9408af37b6f4fa580a cyberbeing gave. If "Mux into a 1920x1080 MKV" is a must, or at lease a catalyzator, then cyberbeing you're very probably right about that the speed-up-big-ass-loading thing I made introduced this bug. I did change the script loading bahavior of vsfilter. And in a MKV file the script is split into many packets, every time VSFilter recieves a script packet, it saves the pack into a temp file, and reload the file. So the loading module will be invoked frequently enough to produce a bug.

[cyberbeing]

If "Mux into a 1920x1080 MKV" is a must, or at least a catalyst, then cyberbeing you're very probably right about that the speed-up-big-ass-loading thing I made introduced this bug. I did change the script loading behavior of vsfilter. And in a MKV file the script is split into many packets, every time VSFilter receives a script packet, it saves the pack into a temp file, and reload the file. So the loading module will be invoked frequently enough to produce a bug.

I haven't actually tested if this only affects MKV files, but it's possible that could be the case. What you said does sound reasonable for how this could be happening. I only wrote the Readme in that way to eliminate variables and hopefully increase the chances of you reproducing the bug. I couldn't exactly send you my computer and upload the entire 16GB MKV to GoogleCode.

[taulin]

i added a comment to the issue on googlecode

[Ekyu]

Hello,

having recently switched to your VSFilter from the threaded one to test it out, I have come across two problems and I'd like to share them with you.

1) Subtitles not showing up

Now this happened so far with only one serie (Kaamelott) which are in .mkv and have normally 3 streams of subtitles (none, french and english).
Subs for this particular release don't show at all while they do with the threaded VSFilter as well as the basic VSFilter (this one obviously being not good enough to handle the CPU heavy subs I have in some releases).

I don't really know which informations I should give you right off the bat about these subs so if you would be kind enough to just point them out to me that would be nice.

2) Weird pixel lines in letters

I have noticed this in the show I'm watching at the moment (Amagami SS http://bakabt.me/159892-amagami-ss-720p-bd-utw.html and potentially in the ED subtitles of Cardcaptor Sakura http://bakabt.me/156241-card-captor-sakura-960x720-blu-ray-flac-coalgirls.html although for Sakura it is only with the threaded VSFilter).

Basically when watching the show some of the letters will have a weird pixel line to them from time to time (it's rare), and if I notice it but I'm too slow to pause to get closer to my TV to check in detail what's that line about and seek back to view that particular sub line again, the line will have disappeared and now everything looks good.
Thus I don't have a screenshot on hand (I'll try to get one though) but let me try to illustrate the problem like so :
Take a big letter like W with it's edges black and the inner part filled in white.
\      \   /     /
  \  _ _ _ _ /
    \  /  \  /
That looks terrible but yeah hope you can get what I mean.
The line of underscore would represent the line of pixels I'm seeing. It can be either here or a bit lower or higher up.

It's really small, doesn't prevent in any way to read the letter, it just looks ugly and I'm the meticulous person searching hours for dead pixels on any new TV / screen I buy so I thought I would just share this to you see if it is something normal that's screwed with the rendering or if it is just me.

Same as with gommorah I'm available for any testing you need and to further comment on these 2 issues.
And last but not least thank you for this moded VSFilter, it's really great and I hope to see it developed enough to be included in an official CCCP release as subs are getting out of hand these days.

[x_xy_y2]

Hello,

having recently switched to your VSFilter from the threaded one to test it out, I have come across two problems and I'd like to share them with you.

1) Subtitles not showing up

Now this happened so far with only one serie (Kaamelott) which are in .mkv and have normally 3 streams of subtitles (none, french and english).
Subs for this particular release don't show at all while they do with the threaded VSFilter as well as the basic VSFilter (this one obviously being not good enough to handle the CPU heavy subs I have in some releases).

I don't really know which informations I should give you right off the bat about these subs so if you would be kind enough to just point them out to me that would be nice.

2) Weird pixel lines in letters

I have noticed this in the show I'm watching at the moment (Amagami SS http://bakabt.me/159892-amagami-ss-720p-bd-utw.html and potentially in the ED subtitles of Cardcaptor Sakura http://bakabt.me/156241-card-captor-sakura-960x720-blu-ray-flac-coalgirls.html although for Sakura it is only with the threaded VSFilter).

Basically when watching the show some of the letters will have a weird pixel line to them from time to time (it's rare), and if I notice it but I'm too slow to pause to get closer to my TV to check in detail what's that line about and seek back to view that particular sub line again, the line will have disappeared and now everything looks good.
Thus I don't have a screenshot on hand (I'll try to get one though) but let me try to illustrate the problem like so :
Take a big letter like W with it's edges black and the inner part filled in white.
\      \   /     /
  \  _ _ _ _ /
    \  /  \  /
That looks terrible but yeah hope you can get what I mean.
The line of underscore would represent the line of pixels I'm seeing. It can be either here or a bit lower or higher up.

It's really small, doesn't prevent in any way to read the letter, it just looks ugly and I'm the meticulous person searching hours for dead pixels on any new TV / screen I buy so I thought I would just share this to you see if it is something normal that's screwed with the rendering or if it is just me.

Same as with gommorah I'm available for any testing you need and to further comment on these 2 issues.
And last but not least thank you for this moded VSFilter, it's really great and I hope to see it developed enough to be included in an official CCCP release as subs are getting out of hand these days.

Pretty good illustration and I got what the problems are. To fix them, first I have to know what they are, which I think I've got via your illustration. Then I need to reproduce them. Most of the time I need the related scripts or even video clips, as simple as possible. And if it can only be reproduced following specific steps, which steps should I follow?  Models like http://code.google.com/p/xy-vsfilter/issues/detail?id=5 and http://code.google.com/p/xy-vsfilter/issues/detail?id=15 are nice.

[JEEB]

Very interesting modifications here. Also, you could've bumped the thread with a post every time you make updates, as I had missed the updates completely until someone mentioned it around

Given that this version doesn't need NT6, some of the modifications could be later presented to the MPC-HC VSFilter repository. With all due respect to gommorrah's work (as he seemingly has less regressions at the moment), NT5 will still be relatively relevant to support for a few more years.

What I see as problems at the moment:

• Not based on current vsfilter code (can be dealt with by rebasing and modifying the patches already in -- git rebase ftw).
• Possible regressions (can be dealt with by having enough testing and having someone go over the patches).

Anyways, good job overall.

Also, what renderers could take in 16bit stuff? I know 10bit is (generally) less supported, but I really haven't tested this stuff. And yes, I know that 10bit is actually 10bit-in-16bit, but still :3

[cyberbeing]

Arguably VSFilter 2.39.x is current stable VSFilter code and everything else are forks including VSFilter 2.40.x (which is a fork of VSFilterMod which was a fork of 2.39.x).

I believe he said rebasing wasn't easily possible because the VSFilter 2.40.x (VSFilterMod fork of a fork) codebase was too different from VSFilter 2.39.x.  He does have adding the VSFilterMod tags on his to-do list, though I expect he needs to do it the hard way with manual patch-work. That hard work will probably pay off if the resulting codebase ends up as a much cleaner port from 2.39.x than the current 2.40.x codebase as a result.

In many ways I could see scrapping dead-slow VSFilter 2.40.x and never looking back once those new tags are added and regressions stomped, while having the MPC-HC project take it on as VSFilter 2.41.x or something.

As for 10-bit & 16-bit, only madVR accepts P010 & P016 at the moment.

[JEEB]

Arguably VSFilter 2.39.x is current stable VSFilter code and everything else are forks including VSFilter 2.40.x (which is a fork of VSFilterMod which was a fork of 2.39.x).

Just noting here that just the fact of X being a fork of Y doesn't mean that X is less stable in any way. Also, I think the version that ended up in MPC-HC's repository got other updates before having stuff merged from VSFilterMod.

This of course doesn't mean that I am saying here that the MPC-HC's current status is better than... actually I'm not sure what is supposed to be "2.39.x", Gulverkli2? Anyhow, just noting something here.

I believe he said rebasing wasn't easily possible because the VSFilter 2.40.x (VSFilterMod fork of a fork) codebase was too different from VSFilter 2.39.x.

Nowhere did I say that putting stuff up-to-date'ish would be easy.

He does have adding the VSFilterMod tags on his to-do list, though I expect he needs to do it the hard way with manual patch-work.

The VSFilterMod-specific stuff isn't in or at least activated in the source code that's in MPC-HC's repo. And it is the general consensus that those tags shouldn't be added to any kind of VSFilter version that is supposed to be used for playback. Let's not make the ASS spec any more herp derp than it already is. (related Aegisub blog post)

In many ways I could see scrapping dead-slow VSFilter 2.40.x and never looking back once those new tags are added and regressions stomped, while having the MPC-HC project take it on as VSFilter 2.41.x or something.

That will depend on how things go on, I would at least like to see a multithreaded renderer stabilized, and then put into the MPC-HC repo as that is the one that is used by Aegisub and related parties.

As for 10-bit & 16-bit, only madVR accepts P010 & P016 at the moment.

Derp.

[cyberbeing]

This of course doesn't mean that I am saying here that the MPC-HC's current status is better than... actually I'm not sure what is supposed to be "2.39.x", Gulverkli2? Anyhow, just noting something here.

Gulverkli2

The VSFilterMod-specific stuff isn't in or at least activated in the source code that's in MPC-HC's repo. And it is the general consensus that those tags shouldn't be added to any kind of VSFilter version that is supposed to be used for playback. Let's not make the ASS spec any more herp derp than it already is. (related Aegisub blog post)

I agree, but are you positive those tags aren't active in 2.40.x? I thought there were, and that was the entire reason the MPC-HC guys merged it into their project and even CCCP started using it (in case idiots used those tags). If they aren't active, then that makes things easy since there would be no rebase or significant feature patching needed. I fully support the idea of not adding those tags to xy-VSFilter if that's the case, as that has been my stance on the issue for a long time.

That will depend on how things go on, I would at least like to see a multithreaded renderer stabilized, and then put into the MPC-HC repo as that is the one that is used by Aegisub and related parties.

Well before you look at threaded-VSFilter seriously, you'd probably want to revert his threaded blur code since it's a major behavior change from regular VSFilter. Though I do have a sense that the MPC-HC guys wouldn't mind just dropping legacy support and going that route.