Add support to the japanese version with undub, and support for audio files in .vx format

[leomontenegro6]

Hi all. I'm Solid One, a brazilian romhacker who made a PT-BR translation of this game in 2017, alongside with the Mega Man ZX Undub hack that brings the american localization to the japanese ROM to restore cut content like the voice clips during cutscenes. A colleague of mine showed this editor, and I'm genuinely surprised to see you managed to edit ActImagine videos with it.

Initially, when we first did the translation, we missed greatly the japanese audio that were cut, so we came up with the idea of creating a functional undub patch that we could combine with our translation. Not only we brought the american assets to the japanese ROM but we also managed to bring back the routines from the european version that renders video subtitles for people to understand the japanese conversations. But one thing we couldn't do back then is editing video and audio files from cutscenes, since they're in a proprietary format (ActImagine) and there's no opensource tool that allows editing them AFAIK.

A few years later, back in 2021, a colleague of mine released a Mega Man ZX PT-BR fandub project, that currently works only in the Legacy Collection, where he reunited a great team of fandubbers, directed them, and managed to edit all video and audio files from cutscenes. The final result was great, but we couldn't port it to the DS version so far. I always wanted to review my translation to include the PT-BR fandub to it, and maybe even the english Fandub mod that was released back in 2023. But that wasn't possible yet due to technical limitations regarding the proprietary video/audio formats and the lack of tools to edit them.

That said, a great feature request I'd like to request for this editor is the possibility of making it become compatible with the japanese version so that we could make use of the cut content from the undub version, and editing the audio files in ".vx" extension, inside the "vx" folder. There are hundreds of files named as "aXXXXX.vx" or "vXXXXX.vx", and if we could edit them somehow, would allow for us to do the porting work. What do you think about this idea?

[AlaryVanEeckhout]

Hello! Sorry for the late response.

It is important to note that the ActImagine library that I am using for my editor (mainly developped by CharlesVanEeckhout) is still far from finished, and while encoding and decoding the audio of vx files is planned, it is not easy to implement due to there being little documentation available for this format, so there's a lot of trial and error and comparing with an obscure version of ffmpeg in the form of a linux executable that can decode vx video and audio. Even the current video encoding strategy is not super effective and is more of a proof of concept. Only the video decoding really works as intended, and that's thanks to an amalgamation of abandoned ffmpeg patches to the source code to support decoding vx files as a list of pngs (that did not support audio or encoding), making it much easier to replicate the process.

With the latest update to the library, some work was done on the audio aspect, but the decoding part is still unusable in its current state, which is why the editor will now likely crash whenever trying to export vx (the library only allows to export both video and audio at once so if one of those fail, the entire process fails).

As for compatibility with the japanese version, is there a specific feature of the editor you noticed did not work only on that version of the game? While it is true that I stated my focus was more on the USA ROM, everything generally works between different versions, and I don't think the vx file format changes depending on the region.

I hadn't realized there were this many vx files in the japanese version compared to the 20 in the USA version, but I guess it makes sense because there is voice acting in each cutscene for both playable characters (which made me realize I might know what causes the voice clips to play in the dialogue files).

In conclusion, your idea is cool, but the editor (or, rather, the ActImagine submodule it uses) is too early in development to allow for such edits. I do not know when we will be able to properly decode and re-encode audio in vx files, especially because we do not have access to the source code of the only ffmpeg version that seems to support vx audio decoding, so the only options left are trial and error or reverse-engineering the vx-related routines of the game, both of which can take a long while.

[leomontenegro6]

Well, maybe we won't need to decode the audio clips. Each aXXXXX.vx has a number associated to it, and I believe we can refer to each audio clip from the script files with a specific audio tag, so if we look at each script block with an audio tag on it, we can probably use it to map each audio file individually, and determine beforehand what each audio clip means.

Take the script for the introduction from Vent, for example: https://github.com/leomontenegro6/megaman-zx-traducao-ptbr/blob/master/scripts/talk_m01_en1.txt#L1

<script_info tipo='Missão 1' personagem='Vent' tamanho='7157' inicio_ponteiros='4' fim_ponteiros='91' inicio_textos='92' />
<#####################>

<p 02><d E903><a 0D><n 03>
Alô! Nós transportamos
seu produto para
qualquer lugar!
<--------------------->

<p 02><d EA03><a 0D><n 03>
Nós somos a transpor-
tadora "Girouette
Expresso"!
<--------------------->

The script above was extracted from the american version and inserted back to the japanese version with a dumper/inserter I wrote in PHP-cli. The <d E903> and <d EA03> tags are used for audio clips, and their values are in hexadecimal and little endian. If you change E903 and EA03 to 03E9 and 03EA, and then convert them to decimal, the result is 1001 and 1002 respectively. And if you check the "vx" folder from the japanese ROM, there are two files named as v01001.vx and v01002.vx, which I assume they should correspond to the dialog blocks above.

That said, since we have all dubbed audio clips in high-quality WAV/OGG from the Legacy Collection version, then if only the encoding part works, that would be enough I guess. In other words, converting a WAV/OGG file to .vx format with the ActImagine library, and then test ingame and see what happens. What do you think?

[AlaryVanEeckhout]

Sorry again for not answering sooner, I forgot about this discussion for a while.

Yeah, I noticed the correlation between the names of the vx files and the 0xF9 special character in the dialogue files since my last message and made a commit in relation to that. It is true that since you are replacing the voicelines with your own audio, you don't need the audio decoder to be functional.

That being said, the problem here is that we have no reference on how to encode audio, so understanding how to at least decode an audio file is the only viable way to understand how encoding it to vx works.

[leomontenegro6]

I understand. Really challenging.

BTW, What about the actimagine submodule repo from CharlesVanEeckhout? Looks like we could minimally convert WAV files to .vx with it. Even though the decoding isn't fully done, what if we simply try generating .vx files using it, and testing ingame? I wonder if that could work, even though the audio quality might not be the same as the original.

[AlaryVanEeckhout]

Yes, the library is somewhat capable of encoding audio, though the result is currently very noisy compared to the original, making it hard to recognize sometimes.