Karaoke Help thread

[mad_cracker]

Hi there^^
Need help with some stuff. I've created a video with alpha channel on AFX so i can use as karaoke. The thing is the opening of the serie never begins at the same time. How can i put the karaoke in the time i want? I tried to do like this: do the karaoke with more frames then the opening and then trimm the karaoke in the begining. Does anyone know how to make it more easyer?
Thanks for the help

[Sylf]

When you layer the afx clip with the OP, use trim() to get the OP clip, layer it, then use the splice function to put the clip back to the whole episode.

[mad_cracker]

Thanks a lot^^
I've tried what you said and it works perfectly^^

[Yazakura]

I'm just wondering about something: Can someone with no past experience learn how to make a karaoke version of a song? And does the file you want to make it from have to be in CD format? Are all the programs you need to download to get this working COMP friendly? (Is this legall in the US?)

Thanks much in advance for the help!

[jfs]

What exactly do you mean by "a karaoke version"?
Usually when I see that, I think of a vocal-less version of the song. If that's what you're after, forget it It's practically impossible to do without having the original track-by-track studio recordings and doing a new mix (leaving out the vocals).
However, if it's a song from a relatively recent anime, there's often vocal-less (off-vocal) versions of them on the singles.

[Soluzar]

Originally Posted by Yazakura

I'm just wondering about something: Can someone with no past experience learn how to make a karaoke version of a song? And does the file you want to make it from have to be in CD format? Are all the programs you need to download to get this working COMP friendly? (Is this legall in the US?)

Lets clarify something. This thread isn't about making versions of songs without a vocal track to sing karaoke to, because that's almost impossible, as JFS stated. It's not possible to entirely remove the singer's voice one it has been mixed in.

This thread is about adding fancy karaoke graphics to the opening credits (OP) of an anime episode. It's a subset of the overall typesetting skillset of anime fansubbing, and simple effects can be done relatively easily, in Aegisub. More complex effects will take you a lot of practice and time to perfect, and might require other software Typesetting is a specialist position within many groups.

My advice to a new fansubber would be to either learn how to do the most basic of Karaoke, which won't take you long, or to just not do any at all. It's not mandatory. I also have to say, if you're worried about legality, then fansubbing is not the hobby for you. It exists because the studios allow it to exist, but it is not legal.

[mad_cracker]

Can anyone tell me how to overlay 3 ou more videos at once. The problem is that the karaoke does never begins at the same time and i have typesets to put in the original video. I wanned to overlay the typesets and the karaoke with the otiginal video at the same time so then i dont have to make the encode twice in the same video^^'
Help plz

[Sylf]

rather than overlay, just use the same encode each time.

concept:
Encode the section with karaoke by itself with the audio track, so you can use it each time. When you encode the episode, delete the karaoke section from the raw, and stick the previously prepared clip in its place.

========================

The Virtualdub method 1:
1. Encode the karaoke clip in the final format. For example, using xvid and mp3 as video and audio codec. Use CBR for the audio.
2. Encode other parts of the episode using the same format as step 1. Use the range selection in virtualdub to set the section you want to encode.
3. Concatenate all the sections you prepared in step 1 & 2 using virtualdub, using "Append AVI segment."
4. Set both video and audio to Direct Stream Copy, and save the whole thing as AVI.
5. For the future episodes, repeat steps 2 thru 4, reusing the clips you prepared in step 1 before.

Probably the easiest way to go for a beginner. It doesn't require much resource, etc, and it works.
========================

The Virtualdub method 2: (not much different from the prev)
1. Encode the karaoke clip using lossless video and audio codec, such as huffyuv and raw PCM.
2. Encode other parts of the episode using the matching codecs as step 1.
3. Concatenate all the sections you prepared in step 1 & 2 using virtualdub, using "Append AVI segment."
4. Set video to Fast Recompress. Set audio to Full Processing Mode. Pick codecs for each. Save the whole thing as AVI.
4.5 Or alternatively, don't encode video and audio together. Encode them separately, and mux the two files together, using other tools, such as AVI Mux GUI.
5. For the future episodes, repeat steps 2 thru 4, reusing the clips you prepared in step 1 before.

This method adds a bit of flexibility, but requires a fair amount of disk space. Encoding the a 24-minute video in lossless format would require up to 4.5 GB, I think.
========================

The AVISynth method: (Just doing the same thing as VDub method 2)
1. Encode the karaoke clip using lossless video and audio codec.
2. Prepare the avisynth script to be used for final encode. Use trim() and alignedsplice() functions (or ++ operators) to put all the videos together
3. Load the script in your favorite encoding engine (e.g. VirtualDub, AVIDemux, MEncoder, x264, MeGUI, etc), and encode.
4. For the future episodes, repeat steps 2 & 3.

This probably is the most preferred method by seasoned encoders who wants to do something like this, since they're using AVISynth to start with anyway. And you can skip the lossless encode of the main episode with this if you really want.

Of course, this requires you to know AVISynth. Which I recommend.

[mad_cracker]

Thanks a lot. I've tried some things you said and i'm starting to get it^^
I'm realy thakful. Here in Portugal we beginers don't have this kind of information. Because ppl here try to make things for themselves and don't like to share it. I think its kind of stupid. If we are fansubs from the same country why not share info to make things get better for our watchers.
Peace and anime^^

[Takeru]

I have been reading this entire thread for the past 3hrs trying to understand some of this. I'm a total n00b at this, I'll admit it.
I really want to figure out how to do this. So if anyone could point me to a very simple tutorial on it or whatever, I'd be very appreciative about it. I have to admit tho, the extent of my programming knowledge is HTML and a slight bit of JavaScript.

[xat]

If you need a basic start, look into Aegisub's documentation. There's a nice typesetting primer which goes over every override you'll be working with when creating karaoke.

[Sylf]

First, you need to learn how to time the karaoke. Many SSA/ASS timing programs, including Substation Alpha, Medusa, and aegisub has this function. (Sabbu 2.7 beta3 thru beta 14 had this function as well.) Basically, you can use \k (immediate fill), \ko (outline fill), and \kf or \K (gradual fill from left to right) with such timing.

Eventually, people start want to have fancier effects. That's where most of this thread and aegisub's tutorial will come handy.


Basic gist of karaoke timing:

Spoiler:

<Zydgardyn> Would you mind explaining how to do Karaoke subs?
<killmer> \k and \K are the commands
<killmer> time the line like normal
<killmer> but dont leave much space at the beginning or end
<killmer> oce you get better at it you can link it easily
<killmer> oce = once
<killmer> \k does highlight then stop
<killmer> \K does highlight continously
<killmer> \K looks the best
<Zydgardyn> do you just put \k at the beginning of the line in the text box (I use SSA)...
<skates316|away> what about the numbers next to the k command ?
<killmer> hehe
<killmer> im getting to that
<killmer> with karaoka do 1 line at a time
<killmer> 1 line will take 1-5 minutes
<killmer> depends on line length and your speed
<killmer> ill explain them again real quick
<killmer> \k = highlight then stop
<killmer> \K = continuous hoighlight
<killmer> i always use \K
<killmer> the karaoke commend will look like this {\K45}
<killmer> the number is the length in centiseconds
<skates316|away> centiseconds?
<killmer> yeah
<killmer> i always call it milliseconds
<killmer> but 1 milli equals .001
<killmer> and ssa reads time in centiseconds which is .01
<killmer> if you do it my way youll never have to figure this out
<killmer> time the line regularly first
<killmer> grab times
<killmer> go back play that row
<killmer> itll leave the brackets
<killmer> which will give you the start time
<killmer> ok
<killmer> here is an example
<killmer> {\K13}Yo{\K10}mi{\K90}ga{\K84}e{\K141}re
<skates316|away> so on 13
<skates316|away> it will start highlighting Yo?
<killmer> {\K13}(13cs highlight->)Yo{\K10}(10cs highlight->)mi{\K90}(90cs highlight->)ga{\K84}(84cs highlight->)e{\K141}(141cs highlight->)re
<killmer> if you wanna link them it pretty easy
<skates316|away> umm on the rave karaoke
<skates316|away> it kinda rolls through
<killmer> yeah
<skates316|away> is that what that line will do?
<killmer> thats the \K command
<skates316|away> okay cool
<killmer> it continuously highlights the line
<killmer> \k highlights real quick and puases for the next \k command
<killmer> always time karaoke by syllable
<killmer> itl never look good by word
<skates316|away> k
<killmer> ok
<killmer> some tips now
<killmer> 1) always do a full line before saving
<killmer> 2) copy the line before entering it just incase
<killmer> 3) do not even just hit enter
<killmer> 4) if it adds time onto the end you fucked up
<Zydgardyn> ... where is the screw-up? is it in the \K times or in the main time?
<killmer> /K times
<killmer> if it ads time your \k's are to short
<killmer> it it cuts time theyre to long
<Zydgardyn> where does it add/cut the time? to the last \k?
<killmer> yep
<killmer> 5) the end of the previous syllable is ALWAYS the start of the next
<killmer> 6) i find sometimes i add a command like {\55} and forget the K
<killmer> if this is done use a calculator
<killmer> centiseconds are 00-99 not 00-60
<killmer> ok
<killmer> now
<killmer> im gonna go into quick linking
<killmer> this is where the previous and current line link
<killmer> this can easily be done by first linking them regularly
<killmer> then adding an extra {\k} or {\K}
<Zydgardyn> How do you link lines regularly?
<killmer> linking is just taking up that space in between lines
<killmer> only link with 1.0 or less
<killmer> so if i wanted to make that sk line link it would look like this
<killmer> ok rules on linking
<killmer> less than 1 second = ok to link
<killmer> more than 1 second = dont do it
<killmer> therefore the most you should have on a link is {/K50}
<killmer> when you link
<killmer> you cut the free time in half
<Zydgardyn> Thanks. I just want to know: is there a command associated with regular linking?
<killmer> no
<killmer> its regular timing
<killmer> ok
<killmer> now how to link karaoke
<killmer> take the free space
<killmer> link it normally like i explained above
<killmer> start at the beginning of your line
<killmer> add an extra \K command before your line
<killmer> and add one at the end if it links with the next line
<killmer> so if i were to link that sk line
<killmer> it would look like this
<killmer> {\K34}{\K13}Yo{\K10}mi{\K90}ga{\K84}e{\K141}re{\K0 }{\K34}
<killmer> that would allow .34 seconds before it actually highlighted the line
<killmer> it would also allow .34 seconds after it finishes highlighting
<killmer> oh and last thing
<killmer> its ALWAYS good plactice to end a word with {\K0}
<killmer> it stops it so it doesnt count the space after as a letter
<killmer> you can really see this with continuous
<killmer> and i think that should do it for todays karaoke lesson
<killmer> that should cover most of karaoke timing

Another guide : Kirika's version

It's a chatlog I received back in 2003. Some of the idea is outdated by now (such as when to use \K etc). But basic idea of how to time karaoke is still valid.

[Thar]

I have some problem with ass code, and hope you guys could help me

I'm new to karaoke, I've maded timing and some simple effect. Really simple: text is in first color, then while singing it's getting filled by the scond color, and while it has already being singed it's getting filled by grey.

So, my problem is with this line:

{\fad(700,300)\pos(320,30)\k100}{\r\t(1100,1500,\1c&H808080&\2c&H808080&)\kf28}ki{\r\t(1380,1780,\1c&H808080&\2c&H808080&)\kf26}mi {\r\t(2040,2420,\1c&H808080&\2c&H808080&)\kf88}wa {\r\t(2520,2920,\1c&H808080&\2c&H808080&)\kf22}{\r\t(2740,3140,\1c&H808080&\2c&H808080&)\kf23}bo{\r\t(2970,3370,\1c&H808080&\2c&H808080&)\kf25}ku {\r\t(3620,3880,\1c&H808080&\2c&H808080&)\kf76}wo {\r\t(3980,4380,\1c&H808080&\2c&H808080&)\kf39}{\r\t(4370,4770,\1c&H808080&\2c&H808080&)\kf17}a{\r\t(4540,4940,\1c&H808080&\2c&H808080&)\kf19}ta{\r\t(5130,7450,\1c&H808080&\2c&H808080&)\kf282}ta{\r\t(7550,7950,\1c&H808080&\2c&H808080&)\kf24}me{\r\t(7790,8190,\1c&H808080&\2c&H808080&)\kf21}te{\r\t(8000,8400,\1c&H808080&\2c&H808080&)\kf59}ru

The syllabe "ru" immediately gets filled by both colours, and remain grey while singing.

What should I do to fix it?

[jfs]

You \t times end earlier than the actual timing of the syllables. That might be the problem. Ie. the fade-to-grey is actually finished before the last syllable is over.

If I add up all the \k times (including the first \k100) I get 849, which is greater than the 8400 in the last \t. Just change the last \t end time to 8590, that should probably fix it.

This shouldn't by any chance be hand-written? (It'd look like the kind of error typical for hand-written kara.)
One could probably argue that it's good for learning, to hand-calculate effects at first, but it's just so error-prone, using a tool to generate kara is just much more effecient ^^;.

[getfresh]

I will have to disagree with you on that on jfs. I have always hand coded my scripts and they say the proof is in the pudding so there it is. Anyways, you can screw up code in an automated scripter just as easy as you can hand coding. The only determining factor is who the operator is.

To fix your code I first started by simplifying it as shown below.

\1c&H808080& <- not needed since it is no longer visible after fill -> *delete from script*

\kf <- just a waste of file space -> use \K *preference

{\r\t(2520,2920,\2c&H808080&)\K22} <- these pauses don't need the \r's or \t's just the \K's or \kf 's-> {\K22}

{\fad(700,300)\pos(320,30)\k100}
{\r\t(1100,1500,\2c&H808080&)\K28}ki
{\r\t(1380,1780,\2c&H808080&)\K26}mi
{\r\t(2040,2420,\2c&H808080&)\K88}wa
{\K22}
{\r\t(2740,3140,\2c&H808080&)\K23}bo
{\r\t(2970,3370,\2c&H808080&)\K25}ku
{\r\t(3620,3880,\2c&H808080&)\K76}wo
{\K39}
{\r\t(4370,4770,\2c&H808080&)\K17}a
{\r\t(4540,4940,\2c&H808080&)\K19}ta
{\r\t(5130,7450,\2c&H808080&)\K282}ta
{\r\t(7550,7950,\2c&H808080&)\K24}me
{\r\t(7790,8190,\2c&H808080&)\K21}te
{\r\t(8000,8400,\2c&H808080&)\K59}ru

Now the code is a bit easier to read.

From what I understand you want a normal \K or \kf fill to happen then after syllable is finished it will grey out. Well jfs was 1/2 right about what the issue is. It _IS_ timing, but not on event effect timing. Its post event best I can tell from what he wrote.

Let’s start with our basic equations


---------------------------------------------------------
group 1

(offset \k + current \K)10 = start time of current \t
start time of current \t + 400 = end time of current \t

incorrect: {\fad(700,300)\pos(320,30)\k100}{\r\t(1100,1500,\2 c&H808080&)\K28}ki
correct: {\fad(700,300)\pos(320,30)\k100\r\t(1280,1680,\2c& H808080&)\K28}ki

---------------------------------------------------------
group 2

start time of previous \t + (current \K)10 = start time of current \t
start time of current \t + 400 = end time of current \t

incorrect: {\r\t(1380,1780,\2c&H808080&)\K26}mi
correct: {\r\t(1540,1940,\2c&H808080&)\K26}mi

---------------------------------------------------------
group 3

start time of previous \t + (current \K)10 = start time of current \t
start time of current \t + 400 = end time of current \t

incorrect: {\r\t(2040,2420,\2c&H808080&)\K88}wa
correct: {\r\t(2420,2820,\2c&H808080&)\K88}wa

---------------------------------------------------------
group 4

start time of previous \t + (current \K)10 = group 4 reference time

{\K22}

---------------------------------------------------------
group 5

group 4 reference time + (current \K)10 = start time of current \t
start time of current \t + 400 = end time of current \t

incorrect: {\r\t(2740,3140,\2c&H808080&)\K23}bo
correct: {\r\t(2870,3270,\2c&H808080&)\K23}bo

---------------------------------------------------------
group 6

start time of previous \t + (current \K)10 = start time of current \t
start time of current \t + 400 = end time of current \t

incorrect: {\r\t(2970,3370,\2c&H808080&)\K25}ku
correct: {\r\t(3120,3520,\2c&H808080&)\K25}ku

---------------------------------------------------------
group 7

start time of previous \t + (current \K)10 = start time of current \t
start time of current \t + 400 = end time of current \t

incorrect: {\r\t(3620,3880,\2c&H808080&)\K76}wo
correct: {\r\t(3880,4280,\2c&H808080&)\K76}wo

---------------------------------------------------------
group 8

start time of previous \t + (current \K)10 = group 8 reference time

{\K39}

---------------------------------------------------------
group 9

group 8 reference time + (current \K)10 = start time of current \t
start time of current \t + 400 = end time of current \t

incorrect: {\r\t(4370,4770,\2c&H808080&)\K17}a
correct: {\r\t(4440,4840,\2c&H808080&)\K17}a

---------------------------------------------------------
group 10

start time of previous \t + (current \K)10 = start time of current \t
start time of current \t + 400 = end time of current \t

incorrect: {\r\t(4540,4940,\2c&H808080&)\K19}ta
correct: {\r\t(4630,5030,\2c&H808080&)\K19}ta

---------------------------------------------------------
group 11

start time of previous \t + (current \K)10 = start time of current \t
start time of current \t + 400 = end time of current \t

incorrect: {\r\t(5130,7450,\2c&H808080&)\K282}ta
correct: {\r\t(7450,7850,\2c&H808080&)\K282}ta

---------------------------------------------------------
group 12

start time of previous \t + (current \K)10 = start time of current \t
start time of current \t + 400 = end time of current \t

incorrect: {\r\t(7550,7950,\2c&H808080&)\K24}me
correct: {\r\t(7690,8090,\2c&H808080&)\K24}me

---------------------------------------------------------
group 13

start time of previous \t + (current \K)10 = start time of current \t
start time of current \t + 400 = end time of current \t

incorrect: {\r\t(7790,8190,\2c&H808080&)\K21}te
correct: {\r\t(7900,8300,\2c&H808080&)\K21}te

---------------------------------------------------------
group 14

start time of previous \t + (current \K)10 = start time of current \t
start time of current \t + \fad time out = end time of current \t

incorrect: {\r\t(8000,8400,\2c&H808080&)\K59}ru
correct: {\r\t(8490,8790,\2c&H808080&)\K59}ru



---------------------------------------------------------
Final product

{\fad(700,300)\pos(320,30)\k100\r\t(1280,1680,\2c& H808080&)\K28}ki{\r\t(1540,1940,\2c&H808080&)\K26} mi {\r\t(2420,2820,\2c&H808080&)\K88}wa {\K22}{\r\t(2870,3270,\2c&H808080&)\K23}bo{\r\t(31 20,3520,\2c&H808080&)\K25}ku {\r\t(3880,4280,\2c&H808080&)\K76}wo {\K39}{\r\t(4440,4840,\2c&H808080&)\K17}a{\r\t(463 0,5030,\2c&H808080&)\K19}ta{\r\t(7450,7850,\2c&H80 8080&)\K282}ta{\r\t(7690,8090,\2c&H808080&)\K24}me {\r\t(7900,8300,\2c&H808080&)\K21}te{\r\t(8490,879 0,\2c&H808080&)\K59}ru

[jfs]

Now I sure wouldn't want to write an effect by hand having ten layers, four rotations and eight clips per syllable... anyway...

On the suject, the question is actually whether the idea is that the syllables should first start fading out after their entire duration has passed, or they should start fading out while still being highlighted. The effect obtained by leaving it as-is and just applying my fix also gives a quite interesting effect, especially if you let color1 and color2 fade towards the same.

(Oh yes, and I prefer \kf for readability, IMO \k and \K just look too similar )

[Shounen]

Or just use this if your a starter.
Should probly explain the most of what timing etc is...
If you want some more tips check this one out.

Then thanks to crumjas handy tools (could help you alot, if you dont wanna add your effects manually)


update:

Originally Posted by jfs

(Oh yes, and I prefer \kf for readability, IMO \k and \K just look too similar )

I second that!

[Sylf]

Originally Posted by Shounen

Or just use this if your a starter.
Should probly explain the most of what timing etc is...
If you want some more tips check this one out.

Then thanks to crumjas handy tools (could help you alot, if you dont wanna add your effects manually)

Next time, read through the thread before making posts like that.
All those tools and links have been posted in this thread already. The first link, especially - I posted that only 5 posts before yours.

[getfresh]

Originally Posted by jfs

Now I sure wouldn't want to write an effect by hand having ten layers, four rotations and eight clips per syllable... anyway...

On the suject, the question is actually whether the idea is that the syllables should first start fading out after their entire duration has passed, or they should start fading out while still being highlighted. The effect obtained by leaving it as-is and just applying my fix also gives a quite interesting effect, especially if you let color1 and color2 fade towards the same.

(Oh yes, and I prefer \kf for readability, IMO \k and \K just look too similar )

The auto scripting apps weren't able when I was active and to this day aren't open ended enough for my liking. And 10 layers? LOL... You are kidding right. The last karaoke I did have around 600 layers per line. And that karaoke was not possible by the auto scripters. I still have not seen it repeated so I don't think anyone has added it.

If you want to do the same old thing everyone else is doing, get yourself an auto scripter. If you want original work, learn the code inside & out, bust open notepad, and get crackin.

[TheFluff]

getfresh: jfs was the person who, AFAIK, invented karaoke-effected furigana. Is that what you call "the same old thing everyone else is doing"?
Also, I suggest you stop going around on forums to brag about your enormous e-penis and instead go back to RTFM some about Aegisub's automation system (which happens to be designed by said jfs). If you claim a full-fledged Lua programming environment that lets you write arbitrary code to do whatever you like to the text can't do what you can do in notepad, you obviously don't have any idea what you're talking about.