Do you do any audio post-processing for your samples?

Discussion in 'UtaHelp' started by delphic, Dec 30, 2018.

  1. delphic

    delphic I don't need you or this town *headbangs* Defender of Defoko

    Likes Received
    Trophy Points
    Hey, I begin with saying that I don't know if I selected the right category for this type of question, if not I apologize!

    I'm making a big Japanese VCV voicebank and I'd like to provide the best quality possible to it with the tools I have. Usually, the only edit I do on my samples is just a simple noise removal, nothing more, that's why I'd like to ask you some advices for audio post-processing, to get a better performance from my recordings so my voicebank can sound better once inputted in UTAU.

    Every reply would help me a lot, thank you!
    6_Jpop CDJapan
  3. WinterdrivE

    WinterdrivE Ritsu's Renegades Defender of Defoko

    Likes Received
    Trophy Points
    Nope. And if you're using Audacity's noise removal, you're probably better off just not. Audacity's noise removal leaves behind more artifacts and degrades the audio more than its worth.

    The reason I don't use any effects (and don't like when other people do) is because mixing is very much done on a song-to-song basis. Whatever EQ you arbitrarily apply to your raw recordings won't fit any given song more than the raw audio would have. And I'd rather not be fighting and trying to undo whatever EQ you chose when I'm mixing my vocals; just give me the raw recordings. Same with compression, especially since compression is even harder to undo and UTAU compresses the heck out of the output by default anyways (which I also don't like, but I digress)

    At most I'd say you can use a high pass filter at a very low frequency to remove any low noise or hum below the vocals, but tbh that's about it. I wouldn't normalize every recording either, which is a suggestion I often see. you end up amplifying each recording by a different amount and it causes mad inconsistencies across the VB.

    Tbh, most problems that people seek to fix with post-processing are often mic quality problems that can be fixed with a better mic. But personally I'd still rather have lower-quality raw recordings than post-processed ones.
    Last edited: Dec 30, 2018
    delphic and Kiyoteru like this.
  4. Avalia-Kasa

    Avalia-Kasa probably a potato tbh Supporter Defender of Defoko

    Likes Received
    Trophy Points
    i usually batch edit them on reaper these days (though my old UTAU liv/layla didn't get that treatment rip in rest you pieces of garbage lmaoo). usually i only use a noise gate, and i do that instead of noise removal because noise removal is basically a fancy EQ that detects noise in a recording & removes and/or lowers the volume on all the frequencies with it.

    the problem with noise removal is that if you batch edit noise removal on the entire bank, there might be different types of noise throughout the bank (like the beginning of your recording session might have high frequency noise but some samples might have lower frequency noise from things like a far-away plane or something similar), so unless you have a ton of noise that's present even in when your vocals come in - in which case you may want to consider a new mic - noise gate is a better bet imo since it doesn't mess with the EQ

    for people who don't know what noise gate is it's basically an insert that removes anything under a certain decibel, so say your noise only pops up around -24dB, just put your noise gate there. there's other features that change the attack/release so that in vcv strings it doesn't sound unnaturally choppy around your consonants as well.

    EQ i only use sparingly in post with voicebanks because i reaally don't want to change anything that UTAU might amplify. the only instance i would is if i've tested it on a single sample beforehand a LOT and find it makes a nicer sound - and it would probably be a boost on some lower frequencies since UTAU tends to amplify the mid-highs.
    delphic likes this.
  5. MillyAqualine

    MillyAqualine Ritsu's Renegades Defender of Defoko

    Likes Received
    Trophy Points
    I rather leave the vocals raw... Not necessarily for mixing purposes, but because in general I prefer having raw vocals, and then adding the modifications once they are rendered ~ Once I did try once an experiment of using noise removal on a sample, play it in UTAU and if you reach a certain level -that is quickly reached in general- and as it was said above, it'll sound highly artificial, kinda like with a tunnel sound

    Other experiment was applying a soft kerovee effect on a bank just for the sake of hearing the result once in UTAU (one of the Testloids I'm doing sometimes)

    The only effect I apply is pitchchange for Kei, since I can't do his voice naturally so I record a depart voice, and then pitch it down (however be careful with pitchchange, may it be higher or lower because just like with the noise removal option, it's easy to mess up)
    delphic likes this.

Share This Page