Friday, 9 November 2012

Review for Peri

After I first posted Peri's download on the UTAforum, one of the administaitors (Halo)  had to validate the post. She PMed me to review Peri. This is what she said:


Okie doke.

Samples: Wow these are like some weird mix between falsetto and belting.
Is there any particular reason you recorded so high, and in a head voice at that? Serious question (having a hard time trying to make it not sound like sarcasm I'm sorry it's 1AM).

You sound like you're struggling and that's never a good thing. Also, it breaks her range. This may as well be a high set.

Secondly, CV samples never have to be more than 2 seconds long; it's actually detrimental to your bank for a few reasons if you hold them out this long;

1) Your pitch wavers, a lot, and UTAU freaks out and tries to force them level: meaning weird barking sounds for PERIANTH.
2) The size of your bank spirals out of control. An average, dual pitch mind you, HQ CV bank is maybe 30MB. This is monopitch, and was 90MB.

Other things to keep in mind: If you want UTAU to handle the samples consistently, you need to record at the same pitch. Keep a guide going in the background so you stay in key while you record.

You should choose a tone you can hold well. The tones you currently recorded at force you up into your head voice, which would be fab for a high/falsetto set, but not so great for a base tone.

Japanese "r" sounds are not English ones! I would advise to use "L" samples instead until you can pronounce the Japanese "r" (sort of a hard, d-like L sound) correctly.

I don't know if it was forced by the falsetto but... The Japanese "a" is an open "ahh" sound.
You have it as almost the a in "cat" and sometimes and "o" in this bank.

You also need to work on having your t, p and k consonants more clicked and less... Well, in English we like to put a puff of air after hard consonants like that. It's more of a "pha" and "khu" in English, but in Japanese it's a fairly clean "ka". Hard to explain how you're supposed to make that sound, but it's worth a try anyway; just mentioning because some of your k files get eaten up.

General Quality: Not great. But you can't really help the mic you have; is it an inbuilt or USB/Webcam?
If you can, grab a Rock Band/Logitech USB Dynamic Mic. They look a little like karaoke mics (don't get the one that looks like a stick in a pad) and cost about $11. Not the best mic in the world but quite a few prominent UTAU voicers use them, and have proved them to be quite the nifty and clear cheap alternative.

Configuration: A little eh... But honestly, I'm not sure I could do better with these waveforms and samples, they're a bit confusing.

All in all, it's the samples causing issues here.

Next time you record, remember:

  • Choose a pitch you can keep steady
  • Only hold it for a little more than a second
  • Japanese Rs
  • Leave a small space, about 100 milliseconds if you use OREMO, in front of each sample for otoing nicely.
  • Make use of your pretty head voice in a high/falsetto set at some point! Not everyone can do that, haha.

She has promise. But it might take a little work.

Also, you might want to alias her for Hiragana if you can, but I guess now that the kana/romaji UST converter is around there's less need.

Sorry if it's a little harsh, I just want to see you, and PERIANTH, improve.

So based on this, I'm going to re-record Peri. Her new VB will be called PERIANTH². I don't know when I'll record her, but I know It will just be a VB identical to her current one, only better. After that I'll start work on her appends.

No comments:

Post a Comment