Karaoke Help thread

[dj_tjerk]

Some people shift it (most people I know do), some encode the OP and ED once, and mux that in. I don't know how sensitive the latter option is to audio artifacts, but shifting the karaoke doesnt make it go out of sync (if you shift it with the right amount :P).

[getfresh]

Originally Posted by seTup

Also, wouldn't cutting a karaoke'd op/ed using a lossless codec only distort and increase the final file size drastically in an unnecessary way?

No... not at all. The reason most groups don't do it is because credits change for the Animation Production staff and the fansub team(if they have credits and hard sub them.). It won't effect the quality at all since hardsub's are rendered to the lossless of a raw then compressed using the desired format. Another reason that it is not done as often as just shifting is because shifting a kara is extremely easy, depending of course on the type of effects used. But for the most part it takes about 1-2 min of time to do a shift. Just find your reference shift point, find your offset or whatever and shift it. The raws these days are pretty much always the same as the last one, even in frame count for the op/ed, every like 4-5 epis you may have a scene bleed of 1 frame.

Also about the fps rate of the video. That will not effect the linear time scale of a subtitle file at all. the subs will still render at the same linear pace you set them to initially. The depth of detail in movement/transformation will just be limited by how many times per second the movements can be captured.

Think of it this way, if you watch something pass in front of you while blinking your eyes over and over, you see less of the movement, but the speed or rate of what has passed in front of you is not effected at all by the how many times in a second you could open and close your eyes, it is rather how smooth the event looks visually. 24 fps or 30 will not matter at all for how fast the subtitles will move, scroll karaoke fills, display effect transformations, etc...

oh, and I'm sure an encoder can correct me if I am wrong but the whole higher framerate is just to smooth out jitter or choppy movement on panning scenes where they are most noticable.

I really don't get who has been teaching you, but you start looking for a new one.

[h3nshin_h3r03s]

Hi guys, im new to the forum but i learn a lot from a while ago. Anyway, I really need everybody's opinion about my problem right now.

I'm sure that i typed my codes correctly but the primary color for the font didnt goes alpha and the outlines didnt change to white properly. Do anybody know why i am going through this? Here is the screenshot

Screenshot

Here is the codes for the syl2end retime part

!retime("syl2end",-$sdur,0)!{\1a&HFF&\an5\be1\pos($scenter,$smiddle)\ 3c&HFFFFFF&}

[jfs]

You're using an old and buggy version of VSFilter. Updating to the newest (2.39, comes with the latest Aegisub and CCCP among other things) will fix the bug and also give you a bunch of new tags to play with.

[h3nshin_h3r03s]

Thanks jfs, I think I got a lot of catching up to do. New tags? I'm feeling left behind here. Well, new tags means more effects ideas, right? Anyway thanks again.

[edogawaconan]

I hope he actually meant 'better/faster sign typeset with .ass'

[vDrag0n]

I know this might not belong here so excuse me.

I have a problem with my Karaoke Encode. I'm using a 120 FPS mp4 RAW with nullframes and i've yet to find a way to remove all the nullframes. First i tried this:

DirectShowSource("D:\Avisynth\Anime.mp4", fps=24, convertfps=true)

it works good, it converts sound and frames. Problem is it removed random frames. That means there are different frames in Opening/Ending each week, making it impossible to shift times accurate. I googled after a solution and tried many different solutions as example i tried to Decimate it by using SelectEvery(5) which was good again but still difference in frame count of the OP/ED. After that i found a plugin for Avisynth called FDecimate which removes frames based on a Threshold, so far so good. It looked like it really could work this way and like i could remove all nullframes. I went to find the metric/threshold number watching alot of frames (*yawn*) after i found that number i encoded. Problem is it totally messed the Video (It removed almost every frame and played a 23 minute video down in 3 minutes).

Does anyone else have a solution to remove all nullframes so i can just shift times of Karaoke each week?

edit: @TheFluff's Top post: "why encode a 120 fps to a 24 fps" because it takes 3-4 hours to encode 120 FPS not to mention nullframes are really not needed. Also nullframes differ each week making the Karaoke not match.

Originally Posted by getfresh

No... not at all. The reason most groups don't do it is because credits change for the Animation Production staff and the fansub team(if they have credits and hard sub them.). It won't effect the quality at all since hardsub's are rendered to the lossless of a raw then compressed using the desired format. Another reason that it is not done as often as just shifting is because shifting a kara is extremely easy, depending of course on the type of effects used. But for the most part it takes about 1-2 min of time to do a shift. Just find your reference shift point, find your offset or whatever and shift it. The raws these days are pretty much always the same as the last one, even in frame count for the op/ed, every like 4-5 epis you may have a scene bleed of 1 frame.

Also about the fps rate of the video. That will not effect the linear time scale of a subtitle file at all. the subs will still render at the same linear pace you set them to initially. The depth of detail in movement/transformation will just be limited by how many times per second the movements can be captured.

Think of it this way, if you watch something pass in front of you while blinking your eyes over and over, you see less of the movement, but the speed or rate of what has passed in front of you is not effected at all by the how many times in a second you could open and close your eyes, it is rather how smooth the event looks visually. 24 fps or 30 will not matter at all for how fast the subtitles will move, scroll karaoke fills, display effect transformations, etc...

oh, and I'm sure an encoder can correct me if I am wrong but the whole higher framerate is just to smooth out jitter or choppy movement on panning scenes where they are most noticable.

I really don't get who has been teaching you, but you start looking for a new one.

it does matter, you should try it . Especially in a OP/ED where the Kanji is shown, when you time the English to the Kanji and next week there is a new RAW with a different in Nullframes (duplicated frames) it will certainly be way off. Another example is time a 120 FPS then convert the FPS to 24 you will also notice a huge difference (i tried it all really).

[TheFluff]

Originally Posted by vDrag0n

words

Karaoke should always be shifted based on audio only (make a shifting line exactly where the first word is sung or where the music starts and shift the entire script based on that) and doesn't have anything to do with video. Sometimes you can use certain points in the video as "anchors" to shift it instead, but since A/V synch on p2p raws isn't all that reliable this is usually a dangerous method, so I would recommend against using it.

Now if you're talking about credits and stuff having to be retimed slightly every week, you're never going to get away from that problem, 120fps or not. The OP/ED is telecined together with the rest of the episode at the studio or TV station every week and there are most likely going to be differences in decimation if nothing else (the OP/ED credits may also change as the series progresses). Using different p2p raw cappers or even the same raw cappers will just further the problem. It has nothing to do with nullframes. If you really really want the OP/ED to stay the same each week, just encode a lossless version of each and use avisynth to splice them into each episode.

Then again I wasn't aware that mp4's could be 120fps, I was pretty sure that mp4box converted that kind of files to VFR, but hey jap cappers have done really odd things before.

Originally Posted by vDrag0n

edit: @TheFluff's Top post: "why encode a 120 fps to a 24 fps" because it takes 3-4 hours to encode 120 FPS not to mention nullframes are really not needed. Also nullframes differ each week making the Karaoke not match.

hurr durr of course I suggested encoding something with 120 coded frames per second, why yes that makes perfect sense

[Tragic]

When hardsubbing karaoke into a file, I am worried that I may unnecessarily be losing quality since the file is gonna be encoded again afterwords for hardsubbing the subtitles.

Is there a way to avoid several encodes?

[Corona]

Do the kara and subs at the same time as opposed to in two different encodes.

Or softsub the script and gain praise from random people over the internet!

[Soichiro]

You could use a lossless encode for the first pass with the karaoke, though that would take a few GB on your hard drive (for the whole episode, only a few hundred MB for just the kara). The best method, however, would be to do all the subtitles, including the kara, at one time, as Corona mentioned.

[mojhh]

Hi
my questions
How can I convert characters to paint ??? ( auto3) or (auto4)




my friend told me that through the program,
but what is the program ??


please anyone helpe me ?

[50119]

this is

the effect used calpospix()

[mojhh]

I mean that through
(auto3) or (auto4)
??

[jfs]

Originally Posted by mojhh

I mean that through
(auto3) or (auto4)
??

No.

Not without major pain on the side of both developing the effect and in rendering it.
(In other words: Don't even try.)

[getfresh]

Originally Posted by TheFluff

Karaoke should always be shifted based on audio only (make a shifting line exactly where the first word is sung or where the music starts and shift the entire script based on that) and doesn't have anything to do with video. Sometimes you can use certain points in the video as "anchors" to shift it instead, but since A/V synch on p2p raws isn't all that reliable this is usually a dangerous method, so I would recommend against using it.

um... Most everyone I know who shifts kara on a regular basis creates a line that is adhered to a scene change and/or keyframe in the op to use as a visual cue for shifting. shifting on audio is much sloppier since the audio should not change ever unless a different version of the op theme is used. In that case your op karaoke needs a different script to begin with, which my possibly use some of the times from the original but will still need adjustments. Anyways, the time sync for the op audio and op video never changes since it is a direct splice from the original clip. If the audio sync has changed the capper fucked up and a different cap should be used or the cap you have should be corrected sync wise since the rest of the audio will be out of sync anyways.

Also syncing by audio will create much more error accumulation than visual syncing any day.

[TheFluff]

Originally Posted by getfresh

um... Most everyone I know who shifts kara on a regular basis creates a line that is adhered to a scene change and/or keyframe in the op to use as a visual cue for shifting. shifting on audio is much sloppier since the audio should not change ever unless a different version of the op theme is used.

the sense is unmaking

How on earth could adjusting audio timing based on the audio itself be more sloppy than adjusting it based on some external cue that may or may not be in synch, and that pretty much always will be off at least +/-40ms because of inverse telecine decimation differences? Even with direct transport stream caps A/V synch can be pretty shoddy and differing not only between stations but also between different episodes on the same station.

Originally Posted by getfresh

Anyways, the time sync for the op audio and op video never changes since it is a direct splice from the original clip.

oh wait, you meant "encode a lossless OP with karaoke, audio and all and splice it in for each ep"? of course now it makes perfect se- ... hey wait a minute if you did that you'd never have to shift the karaoke at all ever unless the OP changes, so you're still not making sense.

Originally Posted by getfresh

Also syncing by audio will create much more error accumulation than visual syncing any day.

i have no idea what the christ you mean with this

[getfresh]

Originally Posted by TheFluff

How on earth could adjusting audio timing based on the audio itself be more sloppy than adjusting it based on some external cue that may or may not be in synch

Well, this is fairly simple to answer. Every time you re-sync by ear you will more than likely be off a bit since you are listening for you "start" instead of using a stable point that can be relied on every episode as a cue up point, aka a scene change.

You can argue your point about "desync of the audio" all day long from raw to raw, but as I have been resyncing scripts since raws were far worse than they are now I think its safe to say that I may know just WTF I'm talking about this time. if you have audio desync of more than a few frames in a raw to begin with you should be fixing the AUDIO sync not adjusting the subs to be out of sync with it.

Originally Posted by TheFluff

oh wait, you meant "encode a lossless OP with karaoke, audio and all and splice it in for each ep"? of course now it makes perfect se- ... hey wait a minute if you did that you'd never have to shift the karaoke at all ever unless the OP changes, so you're still not making sense.

no... sigh. I meant in the CASE that there is a DIFFERENT version of the op or ed used. I.E Fafner used different vocal arrangements of the same ED many times and those had to be retimed to match. Or for instance, Macross frontier and the different lead-ins or intro's into the ending leading up to the main ending section.

Originally Posted by TheFluff

i have no idea what the christ you mean with this

If you don't know what error accumulation is you shouldn't touch timing at all. If you continue to resync to videos with audio desync, AND are actually matching up the lines, you will start acquiring deviation from the base or ORIGINAL based on the differences in hearing each time. With recent raws I have been able to shift tons of scripts with the most deviation being 1 frame randomly on a single line every few episodes using visual cueing. And the audio still is in total sync.

[日本ひきこもり協会]

Well, you obviously time the song to the frames and not the audio. All you need to do is convert the frames to a constant rate. Shifting won't be a problem then.

[TheFluff]

Originally Posted by getfresh

Well, this is fairly simple to answer. Every time you re-sync by ear you will more than likely be off a bit since you are listening for you "start" instead of using a stable point that can be relied on every episode as a cue up point, aka a scene change.

most people just make a shifting line, sync it to the start of the OP (or to some distinct cue), zoom in a lot in aegisub and with that are more accurate than any video cue could ever be (because a lot of the time you don't even listen, you just look at the waveform while zoomed in, grab a time that way and shift based on that); it's really easy to be accurate to within +/-20ms, whereas with video the very best you can get is +/-40ms (with 24fps video)
i know back in your day you didn't have advanced tools like this so maybe that's why you fail to see how easy and accurate it is

Originally Posted by getfresh

If you don't know what error accumulation is you shouldn't touch timing at all. If you continue to resync to videos with audio desync, AND are actually matching up the lines, you will start acquiring deviation from the base or ORIGINAL based on the differences in hearing each time. With recent raws I have been able to shift tons of scripts with the most deviation being 1 frame randomly on a single line every few episodes using visual cueing. And the audio still is in total sync.

I know what it is I just completely fail to see how it could be applicable on this situation
you make a shifting line, all times in the rest of the script are absolute relative to that line, shifting the entire script will not change "error accumulation" in any way, either the script is in synch or it isn't and shifting all lines back or forth by a constant amount of time will not "accumulate errors" in any way whatsoever
this isn't a goddamn amiga, we don't have to deal with ramping or inaccurate clocks


by the way if you're going to play the I AM A VERY EXPERIENCED AND RESPECTED FANSUBBER, I SHOULD KNOW WHAT I'M TALKING ABOUT card, I should mention that on the few occasions I have actually seen audio shifted by video (usually done because the song has hardsubbed japanese lyrics that the entire kara is shifted to) it has always led to perceptible desyncs after like one episode regardless of raw source. "fixing" a/v synch is a really inaccurate business because the only way of doing it is experimentally by ear, and human aural perception is really bad at telling when something is in or out of synch.