Xanadu I and II - General Translation Project(s) Thread

[SamIAm]

We're not going to do any releases until everything is done, if that's what you're asking.

When the day finally comes that we do release the finished translation, however, we'll make it easy for a person to elect not to use the dub. I'll release the script, so you can print it and keep it nearby to read when a cutscene happens if that's what you prefer.

For the two games combined, it will be about 60 pages. 

[SamIAm]

Finally, I've finished my first pass of Chapter 7. I've got 76 savestates archived for this chapter alone.

Also, I went ahead and made a test of me reading my script of the intro for Xanadu I. The script was good. My reading sucked. What can I say? I'm not good at playing wise old men. Too bad there aren't any nerdy middle aged guys in the game.

Finally, for a change of pace, I went and translated a no-frills Japanese walkthrough of Xanadu I to English. Its format is pretty simple:
1. Talk to Person A in (Location)
2. Talk to Person B in (Location)
3. Talk to Person C in (Location)
...etc.

Already, though, I dare say it is the most thorough and accurate walkthrough ever made in English. My plan for now is to release it with the game, hopefully with maps included.

[Arjak]

Holy shit, Sam! I can see the finish line from here! I'm going to need to buy a copy of LoX1 soon! Keep it up!

[Vimtoman]

We're not going to do any releases until everything is done, if that's what you're asking.

When the day finally comes that we do release the finished translation, however, we'll make it easy for a person to elect not to use the dub. I'll release the script, so you can print it and keep it nearby to read when a cutscene happens if that's what you prefer.

For the two games combined, it will be about 60 pages. 

Script sounds good. 
Really looking forward to playing this game now.
How many voice actors do you need.

[SamIAm]

How many voice actors do you need.

Assuming we don't get several super-talents who can convincingly pull off multiple characters, probably about 30.

There are actually about 50 different characters in the games if you include all the bit-parts, but I think we're going to be able to recycle some voices without anyone noticing.

When it comes to casting, I'll be looking for four things, in descending order of importance:
1. That you can pull off a voice that fits the character.
2. That you can act.
3. That your recording quality isn't terrible.
4. That you can accurately check to see whether your take of a line fits the prescribed time, and otherwise export it according to the criteria I set.

This is still weeks away, though, so sit tight. 

[technozombie]

I've always wanted to try voice acting, although I don't think I'll be good at it. I would like to try something where I would have to "change" my voice like trying to sound like an old man or something.

[SamIAm]

Chapter 8 is about 40% tested.

---------------------

This week, I've been sinking huge amounts of time into something that I always knew was going to need it. In Xanadu 2, the "Prologue" that you can choose to view as you start a new game is not a typical cutscene, but is rather the normal game engine on auto-pilot with every line of dialogue being voiced via redbook audio while simultaneously being printed in ordinary text boxes.

What's challenging about translating this is that first of all, the game engine and the CD audio never really check to see where the other one is; they just start at the same time and depend on everything going smoothly. The text boxes have delay variables at the end of them, denoted in frames, and when the game counts off the delay, the next text box is loaded. But there are lots of other factors involved:

1. Both the Japanese and the English translation print at one character per frame. If the Japanese text box has 20 characters and the corresponding English text box has 35 characters, then you need to subtract 15 frames from the delay variable.
2. Many names in the Japanese version are printed using an alternate font that takes one frame to turn on, and one frame to turn off. The English names are not.
3. If you split one text box into two text boxes, you have to factor in six* frames to the English frame count total for the transition.
4. Wrapping to another line takes a frame by itself, except in certain complicated circumstances.
5. For some reason, the original game has lots of delay variables paired. Sometimes, it's necessary when the delay is greater than 255 frames, but often, that's not the case. Instead of one variable with a value of 120, it will have two with values of 60 each. And guess what? The game takes one frame to begin counting down a delay variable, so if you want to use one variable instead of two (which is easier, because you're recalculating all the delay variables to match the English anyway), then you need to add one extra frame. I actually said to hell with it and started splitting my recalculated variables.

There are dozens and dozens of text boxes worth of lines, and just figuring all this crap out took hours. I've done a lot of the recalculating now, but it will still be hours of exhausting work more to get everything truly synced up, which it needs to be. You might not think it's such a big deal if the timing gets off by a frame or two, but since the game and the CD are running separately, little mistakes can compound. It would be very easy for the audio and the text boxes to be de-synced by a half-second by the end, and if all the mistakes are little, it will be very difficult to pinpoint where things got off track.

I'm probably going to have to record footage via Mednafen of the original and the translation, and carefully check one box at a time whether everything is synced. It's a damn mess.

The only bright side is, if we're paying this much attention, it actually won't be that hard to clean up little mistakes that Falcom themselves made. Some of the sound effects are ever-so-slightly not in sync, and I'll be able to improve their timing.

So yeah, that's what I'm doing with myself this week.

*I still have to verify this.

[Johnpv]

Holy shit, that you're taking the time to do that, and do it right, is INSANELY impressive. 

[esteban]

Sam: INSANITY!

[SamIAm]

It's starting to work! Ha ha ha ha!! 

There's some variable that I haven't been able to pin down which is causing some English text boxes to hang around for a few frames longer than they should, but that's OK. I made a video rip of the Japanese game and used VirtualDub to count the frames between each and every text box, recording the results along the way, and now, I'm going through an English video rip and doing corrections. Add a frame here, subtract a frame there...I've already got the first few scenes lining up with the original Japanese exactly. And I do mean exactly. 

There is one more variable that's a minor pain to deal with right now, and also has the potential to bother us again later. Elmer went all out and has a great new compression scheme working in the game that gives me all the space I'll need for text; however, the amount of time it takes the CPU to move around and decompress things has changed. For example, in this prologue, there's a moment when the characters enter a castle, the screen fades to black, and a new block of data is loaded behind the scenes before the castle interior fades in. In the translated game, this transition is taking an extra several frames. For the player, this is only another tenth of a second or so of waiting, but for me, this is another delay that needs to be factored in.

The reason why this might bother us again later is that we're using Mednafen to test all this stuff, and there's a chance that real hardware might take a significantly different amount of time to decompress stuff. Fortunately, there aren't that many transitions where things are loaded like this, and if we're only off by a frame or two each time, it won't be a big deal. However, this is why it's important to be diligent everywhere else. If, toward the end of the scene, we've drifted five frames off, it's not really a problem. If we're drifted twenty or thirty, on the other hand, our translation will look like crap.

Part of me is eager to get this done and get back to editing. Another part of me, though, is happy for the chance to take a break from dealing with words and spend some time playing with my calculator. 

[SamIAm]

The first pass of Xanadu 1 is now done all the way through Chapter 8!

Although my job will be busy through the rest of the month, I am going to attempt to get the rest of the game done before April, or at least to reach Chapter 10.

In other news, the Xanadu 2 prologue text boxes are now fully synced to the voices. There's some kind of factor going on which I don't quite understand that's causing tiny shifts in the real delays of each text box. In the end, when they're all played, the last box will finish within one or two frames of where the last box in the original game finished. However, if I locate which box(es) are off and correct them, it can give rise to desyncronizations in other places. Anyway, these differences are literally imperceptible, and the game functions just fine. Editing will be a bit more of a pain, but we'll get through it.

I'm going to keep taking my time putting the dub scripts together. It will definitely pay off if I make them as good as I can before opening auditions and beginning recording/mixing. If you want to try out, it's probably a good idea to start checking back here around mid-April.

[elmer]

The first pass of Xanadu 1 is now done all the way through Chapter 8!
...
I'm going to keep taking my time putting the dub scripts together. It will definitely pay off if I make them as good as I can before opening auditions and beginning recording/mixing. If you want to try out, it's probably a good idea to start checking back here around mid-April.

We're really getting there! It's just one heck of a long-slow-and-detailed job. 

[Vimtoman]

Guy's you are the Elite.

[wyndcrosser]

congrats Sam. Very interested in this, I played it previously all the way through, but it was difficult with my limited japanese lol

[SamIAm]

Update time.

Chapter 9 of Xanadu 1 is finished, except for literally two or three text boxes that sit behind a weird bug that elmer will hopefully have worked out soon. It happens at a moment when the game does something that it never does anywhere else, so it's neither terribly surprising nor worrying that the same bug will pop up elsewhere.

I have spent the last several days doing pre-production work for the dub. Primarily, this means setting up project files in Audacity with the background music and sound effects totally mixed in, and the positions of every spoken line marked so that when I get actual takes from people, I can very quickly copy/paste them in and produce previews within minutes. I'll need another week or two to finish this process. The real objective is to make it so that I can immediately have all the actors hear what they sound like together, and do retakes to improve the overall performance while their interest is fresh.

I'm also working on doing a complete test-dub using my own voice. It's pretty embarrassing listening to the playback, but it's also teaching me a lot. The English lines are all measured so that they'll fit in the same slot of time as the original Japanese lines, being neither longer nor shorter. However, if I were to just leave it at that, it would lead to another problem: the pauses between the lines would be the same as the original Japanese, and that leads to some unnatural sounding exchanges in places. It's one of those differences between languages that's easy to overlook. Anyway, by putting my own voice in there, it's really easy to hear when the English actually needs to end a half-second earlier or later than the Japanese, and noting that in the script ahead of time is going to save the actors a lot of trouble.

I didn't do either of the above-mentioned steps when I made the first Xanadu II dub five years ago. I was in some kind of big hurry back then. When I finally got everyone's takes in, I had to start on the mixing from square one, which lead to sloppiness, and also to delays that made some people start to lose interest. Also, when playing back the results, even though (most) people had done a really good job following the target-time numbers after every line in the script, the pauses in between certain lines were strange, and I knew it was going to need some major reworking.

Of course, the old project began its first major stall around that time, so I never quite got around to it.

Anyway, it really makes me wonder how professional dubbing studios handle this sort of thing. It seems like if you didn't want a product that sounds totally weird and detached from the video, you would have no choice but to go through these steps before declaring your script finished and bringing in actors. However, it's all very time-consuming, and it would have been relatively difficult with 20+ year old technology. If I had to guess, I'd say they probably print off lots of copies of the draft-scripts, get a few people to sit around a TV or a projector, then repeat the video footage over and over with the sound off, with people test-reading the lines to each other in real time and making notes.

I could be wrong, though.