Resource icon

Jey's OTOing tips and tricks

This is a collection of various ways to look at otoing-
It's an ever-growing/expanding guide of tips, tricks, and various oddities.
Consider this a written version of my mental rule-book.

This is heavily simplified for the masses- If you have any questions you can feel free to ask!
NOTE: This is NOT a "how to oto" tutorial, but more of "ways". This guide ties in with my OTOing: How TO tutorial and I suggest reading that or similar tutorials before reading this!
This is very much a WIP



Value rounding:
For VCV and VC samples I suggest working in multiples of 5 for simplicity.
For CV (including those in a CVVC bank) samples I recommend working in multiples of 1 (any value)



overlap placement:
CV:
For CV style samples, I suggest and overlap range of 30-45.
  • hard unaspirated consonants should sit above 30 and below 35. overlap should sit evenly in the space before the consonant.
  • hard aspirated consonants should sit around 33-36 (lets be real....just set it to 35 and forget) overlap should sit evenly in the space before the consonant.
  • fricative consonants...35...always 35
  • fluid consonants should be at 35-40
  • RULES FOR ABOVE 40: Last resort- if you don't need it then don't do it.
VCV:
For VCV I suggest an overlap range of 70-100.​
VC:
time to break common rules- this is directly related to the 2.0x time preuttence rule.
this is based on half the max achievable preutterence.​



Preutterence placement:
CV/VCV:
People tend to place it directly between the consonant and vowel...no...actually place it ever so slightly into the vowel.
Please refer to this: https://en.wikipedia.org/wiki/Voice-onset_time I will explain further if needed at a later point.​
VC:
place it at the end of the vowel. but have it as high as possible! it sounds weird- but doing this pushes the blend as early as possible and prevents you from hearing it. This also is good for non-japanese banks with accents...​



The game of preutterence:
Preutterence always goes after the overlap. period.
them preutterence defines the space that utau has to perform the overlap between notes, having too little will crush notes and too much can smash the overlap to nothing in congested sections.
How you treat overlap will change based on the specific voicebank style.

Preutterence multiplications:
2.0x:
Want an UTAU voicebank that will never glitch in envelopes? ever? The 2x preutterance rule is a safe-guard rule for uncertain situations. At 2x UTAU will perfectly adjust the overlap and envelope glitches within that note should be nearly impossible.
ex: 200 overlap and 400 preuttence will never cause crossfade issues with the preceding note. 85 and 170. etc.
Whether the overlap is 80 or 80,000 as long as the preuttence is double you are safe.​
1.5x:
1.5x the overlap is the crossing point, at this point the allocated overlap space is the largest. at 1.5x the overlap you can encounter the overlap of the current note continuing beyond that of the previous note. A minor risk of envelope issues but generally not anything to worry about. While glitching may not be noticeable, it can effect the the render in a way that sounds unnatural.
ex: 85 overlap and 127 preutterence.​
1.4x:
1.4x is my personal limit- below this the risk of crossfacding issues is much too high.
At this point the crossfade space has begun to shrink again but the risk of the envelope not adjusting/fitting properly has increased.
ex: 85 overlap and 119 preuttence (rounded up to 120)​
1.0x:
commonly seen, having the overlap and preuttence the same will more often that not have crossfade issues.
0.5x: also common, this is done to single vowels for timing but isn't recommended as it will make the overlap go backwards into itself.​

rounding recommendations: round up to a multiple of five or down to a multiple of one.
above 2.0x: no rounding
above 1.5x: round up
below 1.5x: round up​

Method specific:
CV samples:
Generally, the low overlaps values make preutterence placement less of an issue. most of the time the consonants length alone will means the preutterence will be in a good place.
hard consonants: generally the 2.0x applies and should be no issue with the minimum overlap and consonants length.
Fricatives: the length of the consonants should cause it to meet the 2.0x requirement. max overlap below 115 to avoid issues stated previously.
fluid consonants: I suggest a minimum preutterence of 80-85. keeping these fast (but still audible) sounds more natural then having them too long.
single vowels:
use the 1.4x rule for vowels to keep the timing reasonable.​
VCV:
The length should keep the preutterence reasonable. Use the 2.0x rule for Y and W consonants.​
VC:
Use the 2.0x rule.​


Author
na4a4a
Views
413
First release
Last update
Rating
5.00 star(s) 1 ratings

More resources from na4a4a