cv otoing question!

[neopronounz]

so i have my very first utau, and im mostly done otoing him !! he enunciates properly now after i completely overhauled them and actually followed a tutorial, but i'm looking to clean them up a bit more!! he has a hard time transitioning between notes when they overlap and just kind of sounds like hes mumbling. i know since hes a cv hes going to sound choppy either way, but i want to make it a bit smoother and have words be more cohesive if i can!

the only tip ive seen for this was done in setParam, and all it said was to put the overlap where the vowel WOULD end on a vcv voicebank, which is too vague for me to conceptualize on my own ^^;; are there any other guides with visuals for something like that, or other methods to do it?

 

[Kiyoteru]

Here's a guide that you can refer to for CV otoing: https://utaforum.net/resources/cv-otoing-guide.269/

 

[neopronounz]

Here's a guide that you can refer to for CV otoing: https://utaforum.net/resources/cv-otoing-guide.269/

this is actually the guide i used when i overhauled the otos !!! it was super helpful in getting him to enunciate better on short fast notes, but he still has trouble with notes meant to fade into each other :c so thats what im asking about !!! thank you still though!!! ^^

 

[수연 <Suyeon>]

CV libraries are going to have transitional problems with vowels because of how they're made. If you want cleaner transitions, you're going to have to either record vowel-vowel transitions or use the built-in Crossfade tool (in UTAU, Alt+T, Alt+U, Alt+C in that order).

 

[dead_byte]

CV voicebanks are choppy in general. Granted, you already know that, but there's no particular method to make them smoother. It honestly just takes experience and gaining a more intuitive understanding of how the parameters work. Most significantly, I would suggest experimenting with the Overlap of samples, particularly any samples that don't contain plosives or affricates ( so p, b, t, d, k, g would be plosives, and ch, ts, and sometimes j and z depending on how you pronounce those sounds in Japanese are affricates. ), since the Overlap value relative to the Preutterance, is primarily what determines how the crossfade is done in UTAU. I also might recommend experimenting with a "rentanjutsu" recording method. It's a type of recording method for CV ( and sometimes CVVC ) where the samples are recorded in "strings" like "a ka ki ku ke ko" and so forth, not unlike how VCV is done. The idea for rentanjutsu-style CV is that they will sound more natural with other sounds around them, as opposed to being isolated samples like in regular CV. They generally sound better for CV voicebanks, if only a tiny bit.