A new way of doing Arpasing English maybe?

Discussion in 'UTAU Discussion' started by PeriodicalUtau, Sep 11, 2017.

  1. PeriodicalUtau

    PeriodicalUtau Teto's Territory

    Messages
    71
    Likes Received
    15
    Trophy Points
    17
    So I don't have any samples or anything, but I've just thought of doing something a little different with Arpasing. So in VCCV (to my knowledge) you have stuff that has 2 consonants in front of the vowel, we also have that in Arpasing, but it's a little far apart in the individual recordings when done "correctly" (even though there really is no "correct" way to do it as long as it works lol). My "totally original" idea stems from VCCV in the sense that the recordings have almost no distance between the consonants and we would now input notes something like
    th ah/ ah s/ s ah/ ah ng/ g ih/ ih s/ st ow/ ow t/ t ih/ ih l/ l iy / iy k/ k uw/ uw l/ l -/ REST
    (the song is totally cool)
    as opposed to |this right here depends on how you recorded ya bank
    \/
    th ah/ ah s/ s ah/ ah n/ n g/ g ih/ ih s/ s t/ t ow/ ow t/ t ih/ ih l/ l iy/ iy k/ k uw/ uw l/ REST
    (the sang is totally cool)

    In theory it would work, but I have no experience recording arpasing banks so idk fam. It really depends on how much extra-non-moresampler otoing you want to do? Moresampler already does produce pretty great otoes for basic banks (OTOES AS A BASE TO START FROM) and such, so idk if it's capable of doing this on it's own, again I haven't tried it.
     
  2.  
  3. Kiyoteru

    Kiyoteru Local Sensei / Kuya Supporter Defender of Defoko

    Messages
    2,190
    Likes Received
    3,078
    Trophy Points
    157
    This is a similar discussion to one I've seen on Twitter a while ago, however I don't think I'll be able to find the exact tweets. I'll rephrase what I said then.

    Arpasing is strictly diphone based. It's true that in some cases, consonant clusters sound awkwardly timed and weirdly pronounced. However, if you have enough redundant samples, then the Arpasing Assistant should be able to select a particular sample that works best for the context.

    For example, the word "start". You might have two samples recorded "ta" and "sta" and they both produce a "t aa" OTO. However, one of them would be named "t aa" and the other one would be named "t aa1". Because the assistant depends on context, it will choose the "t aa" sample that came from the "sta" recording for the word "start", which results in a much more natural "st" consonant cluster.

    Any OTO entry that is non-standard will cause Arpasing Assistant to cease functioning.
     
    PeriodicalUtau likes this.
  4. PeriodicalUtau

    PeriodicalUtau Teto's Territory

    Messages
    71
    Likes Received
    15
    Trophy Points
    17
    *doesn't use arpasing assistant because I like doin it by hand* but in theory, would it work?
     
  5. Kiyoteru

    Kiyoteru Local Sensei / Kuya Supporter Defender of Defoko

    Messages
    2,190
    Likes Received
    3,078
    Trophy Points
    157
    At that point, it would no longer be Arpasing. With the release of the 0.2.0 reclist, having higher diphone redundancy was chosen over including triphone (or more) samples. IIRC it was too unwieldy to take that approach. If you want to know more, you'll have better chances of communicating with Kanru on VocalSynth Space or on Twitter. There is currently a thread about the development of the next reclist: https://vocalsynth.space/d/62-help-me-with-revising-the-experimental-arpasing-recording-script/
     
    PeriodicalUtau likes this.

Share This Page