...recording advice please?

Nohkara

Pronouns: He/him
Supporter
Defender of Defoko
So

When I record VCV or CVVC Japanese I get very annoying problem:

When I record in my the most comfort range (A3 to D4), my recordings looks literally like this - like I mean every time when I really try. The volumes of vowels does change dramatically. I do record with rode procast microphone and recording in my big room if that matters.

Screen Shot 2016-04-24 at 18.10.07.png

When I do record super low and calm possible (which is not easy like at G#3) recordings looks like this.

Screen Shot 2016-04-24 at 18.09.37.png

My low recordings has, of course, bit more bg noises (and sometimes get super weird "noises/error(?)" like in this sample at "fu") than my higher recordings BUT lower voice sounds may more "rich" that higher recordings (and I found it also much easier to mix/use :'D).

Is there any way to me to "fix" this problem? I want that my normal range/higher range recordings to be better (like in lower one), so I could do a GOOD multi pitch some day.

Also, would be good to know what causes that weird "error noise" like thing when I record at very low. >.< It's not coming always but sometimes (Was I too close to mic? Was mic set to be too sensitive?).
 

Kiyoteru

UtaForum power user
Supporter
Defender of Defoko
You're right, that's an EXTREMELY dramatic volume change. Usually vowels will vary in volume slightly, just because of the formant content (while sounding at pretty much the same level). It could be a factor of pronunciation. Something else to try could be compression? You can do batch processing with Audacity, so you could run a compressor on a whole voicebank.
Most resamplers (including UTAU-Synth's engine) normalize samples anyway, so maybe it'll be okay in the end.
 
  • Like
Reactions: GothAmaterasu

Nohkara

Pronouns: He/him
Supporter
Defender of Defoko
Thread starter
You're right, that's an EXTREMELY dramatic volume change. Usually vowels will vary in volume slightly, just because of the formant content (while sounding at pretty much the same level). It could be a factor of pronunciation. Something else to try could be compression? You can do batch processing with Audacity, so you could run a compressor on a whole voicebank.
Most resamplers (including UTAU-Synth's engine) normalize samples anyway, so maybe it'll be okay in the end.
How do I "batch processing" in Audacity (I don't use it much so IDK... ^^; )
 

na4a4a

Outwardly Opinionated and Harshly Critical
Supporter
Defender of Defoko
Sometimes volume changes appear more drastic but it depends on the sound you are making and the pitch in which you are making it.
Example: the "n" sound at a high pitch will appear much smaller than at a low pitch. But at the pitch you demonstrated that shouldn't be the case.
You may just be recording at a pitch that you are uncomfortable/too unstable at and getting a weak sound because of it.

The noise/distortion may of been triggered by the fact f/h are unvoiced and thus have to pitch. You can edit the frq map and remove the pitch curve over those consonants.
Also you should try using the old resampler from utau 2.76 (resampler version 10) or speedwagon's world generation to generate your frq files.

You can get the previous resampler in the old 2.76 zip file on the utau website. Be sure to rename it from resampler.exe to reampler10.exe so you don't delete the newer resampler.

You can also get SpeedWagon from http://custom-made.seesaa.net/s/article/312531314.html. Just launch SpeedWagon_DandD.exe and drag you wav files into the small window. If should be relatively fast.
You can also use the other "resampler" exe in conjunction with a copy of resampler to generate your frq files with that instead.
Please note that you may need to fix the resampler cmd file for that to work though...
 

Jeffrey

local hiatus'er
Defender of Defoko
I think it's the mic position. Based on my simple analysis, you positioned your mic where vowels like U and I (don't take this out of context, please) have their sounds obstructed by your lips, cheek or chin. You must've had your mic way behind your mouth, or all of my simple analysis is wrong.

I (kind of) pronounce stringed recordings similarly and they don't increase volume like that. I've had my mic right beside my mouth. A little geometry and analysis could really help. Cloth or Cotton around your mic could help.

EDIT: Wavelength and Decibels might or might not align. Mostly not aligned for me.

Useless diagram:
Capture.JPG
 

Similar threads