DeepVocal Voicebank Creation Tutorial - Installing and Testing Your Voicebank
Installation
This part isn’t too complicated! Simply copy the packaged voicebank (the folder that contains SKC, SKI, and voice.sksd) into the editor’s singers folder. This path will be C:\Program Files (x86)\DeepVocal\singers, unless you changed the installation path while you were installing the DeepVocal editor. Since the Program Files (x86) folder is protected by Windows, you’ll need to allow the files to move there (this doesn’t take long because there are only a few of them). You’re now done installing the voicebank! If the editor is open, make sure to restart it.
Testing
In order to test your voicebank, you’re going to need to make or open a .dv sequence file. These act the same as an UTAU .ust, OpenUtau .ustx, or .mid file and contain timing, note, lyric, and track data for a song.
The easiest way to do this would be to find a .dv file online, but these are very uncommon. Instead, I’ve provided this one for you! Before using it, make sure to right-click the track at the top and set the singer to the one you just made.
If you’re converting from a UST or something, I recommend using UtaFormatix. Drag the sequence file you have, then scroll nearly to the bottom of the “Select Output Format” page. Select Dv. It will now show you a window where you can convert a bunch of parameters. Since your singer doesn’t have a singer dictionary (which is different than the phonetic dictionary, which you can learn about on the next page of the tutorial), make sure that (under “Cleanup and convert lyrics) the target lyrics type is Romaji CV. Make sure that “Convert lyrics in Chinese […]” is off, since this is a Japanese voicebank. Under “Replace Lyrics”, add another replacement rule. Set the “filter type” to “exact”, then the “filter” to “n”, “match type” to “All”, and “to” to “N”. This will convert the romaji syllabic n to N, the same way it is in the voicebank. Leave all other settings at default, but turn off “Convert Pitch Parameters”. You can experiment with this on your own, but the way DeepVocal handles pitch is quite different from UTAU, so be warned that it may sound weird. Click “NEXT” and then download the sequence.
Thank you for following my tutorial! The rest of this process is optional, so if you’re happy with how your singer is you can stop here and make songs!
Next Step (Optional): Creating a Dictionary File