Straightforward, just some tips here. I'm no expert, but, ehh I'll post a thing anyway. Things are subject to change and things will be added/removed. This was not professionally made.
- When recording, do not leave much space between words, like at all! That's bad. Record like a vcv bank. (I've been told to just use constant velocity if you like recording slower and whatever so, just do that I guess)
- When USTing, VC notes should not go over 60 or 120 milliseconds, this is just like how VCCV works.
- These buttons
- Ok so you can have multipitch in your arpasing bank, but you have to do it this way for now, with a pitch in the main folder and other pitches in sub folders:
- I can't tune, but if I knew how, I'd recommend you tune like you tune VCCV English. It works just the same.
- When recording, your samples don't need to be absurdly long, but if they are, like I said in my first bullet, DO NOT LEAVE A LOT OF SPACE BETWEEN YOUR WORDS
- When recording, you can sing or talk and it'll work out pretty decent anyway. Monitoring your pronounciation helps though if you're singing.
- When generating your oto for arpasing in moresampler, you must have your index.csv in the same folder as your wavs. This counts for multipitch as well. Index.csv in every pitch.
- Male arpasing voicebanks like TIPS better in my experience. Pretty random, but it just works better.
- Careful when using usts made by other people, as if they are already parsed into the arpasing method, using them as plug and play could be a real challenge my friend, I'm sorry.
- The default reclists that are released from Kanru Hua are bound to be missing phonemes. The method he uses to create the list only accounts for the most used words in songs. You can make your own lists or add to the existing list to accomodate for missing phonemes.
- Arpasing is a method, a standard to which English reclists can be made on based on Arpabet and be understood without having to learn a completely new phoneme system for every different reclist.
Last edited: