Event IDs started at 1 in the new version rather than 0 in the old version.The events used slightly different numbers in the “event number” column.The new VSQ removed information pertaining to the harmony tracks in the original.The results were more dramatic than I expected, as it showed hundreds of differences. Thankfully, the internet came through with a difference checker that would work. The ASCII control characters made copying and pasting the text into an online difference checker impossible. Both of them were hardcoded as read-only, meaning I couldn’t edit them even if I saved the file under a different name.
The new file was around 4000 lines shorter than the old one, and the shorter file was over 8000 lines long. The first thing I notice is a difference in length.
#Editra wont open free#
I opened both of the files in a free program called Editra to compare them. The black characters are ASCII control characters. Editra is a free program for coding many many file types. It was a problem with encoding! The original creator of the VSQ conversion tool probably ignored the encoding differences because they didn’t affect how Vocaloid read the files, but when MMD read them, it would get confused and the VSQ files make MMD crash. The old file didn’t crash MMD, but the converted file did. I opened the “Your Gonna Go Far Kid” VSQ in Vocaloid 4 and sent it through the converter. If the phonemes aren’t the problem, what could it be?Įvery VSQ file is actually an INI file with certain information in it, which means, among other things, that the files can be read as text. I opened up the text for “Your Gonna Go Far Kid,” and it said the phonemes were in English. The text file tells you what the phonemes are on each note. I remembered that MMD creates a text file every time it finishes processing a VSQ file. The following article is brought to you by this screen. I made a short file in English, converted the phonemes to the Japanese phoneme system, and put it into MMD. It was time to go test the phoneme theory. One was for a Japanese song, Melody, and the other was for an English song, a cover of “ You’re Gonna Go Far Kid“.
I decided to test the VSQ motion data creation process with two VSQ files that downloaded. If the problem was a phoneme reading issue, I could probably extend the MMD phoneme library to use the English phonemes… somehow. There was a theory that the crash was caused because MMD can’t read English-language Vocaloid phonemes. There is a converter, but without re-saving it in Vocaliod 1, 2, or using some other work-around, the VSQ files make MMD crash. Unfortunately, I had Vocaliod 4, which creates VSQX files, which MMD can’t read. I wanted to use the Vocaloid file I had made to make an MMD lip sync. If you do try it, make a backup of your files before you do any testing. Please don’t try this if you don’t have experience with coding. This kind of messing about can be damaging to your computer. I would have liked to provide a solution, but I didn’t quite have the technical skills to make it work. The technical, detailed explanation of how VSQ encoding can cause two files of the same song to produce different results is below. These tricks should correct the VSQ files make MMD crash. The way to fix this issue now is to modify the VSQ file, and the methods to do so are covered in other articles, like Making VSQ Files Without Vocaloid and Exporting VSQX to VSQ. The English-speaker would think the word means “a feeling of distress”, while a French-speaker would think it means “a food staple”. If you were to show the word “pain” to somebody who only spoke English and someone else who only spoke French, they would both understand that it is a word. Why? The simple answer is “encoding.” The two programs are reading the same files, but they don’t get the same information. Inputting Japanese phonemes would give the correct lip sync. FAL only recognizes Japanese phonemes, so the only notes that show up are those with the phoneme “e”, produced by the sound “eh”, which is used in both languages. This is evident if you have ever put an English song into Face and Lips. MMD and FAL have different ways of interpreting the VSQ data, and these differences result in the same input data generating different results. FAL can refuse to display notes that MMD can understand, and MMD crashes when attempting to process certain files. On the surface, it might look like the programs work the exact same way, but if you have spent enough time putting VSQs into both programs, you might realize that there are some important differences. MMD and Face and Lips are two different programs that use Vocaloid VSQ files to create lip sync for MMD. Why do these VSQ files make MMD crash? Why does FAL not show all of the notes in a song? How do MMD and FAL read VSQ files?
0 kommentar(er)
