SynthV Synthesizer V Studio 2
- Thread starter lIlI
- Start date
I was wrong. There absolutely is "engine noise" in SVS2. It exists roughly between 5 kHz and 10 kHz, and manifests itself as a prickly metallic sound.
This can be isolated (and removed) using the spectral dynamics feature in Pro-Q 4, but doing so also affects "sh" and "ch" sounds.
Like most unwanted noise, it is much more noticeable on pitched sounds (AKA vowels); and more noticeable on higher notes, and/or vocal databases that have more upper harmonics (AKA audible vowel harmonics that traverse across that 5 to 10 kHz noise zone).
Importantly, this noise is far far far more noticeable on SVS1 voice databases, although it still exists with SVS2 voices.
If anyone's interested in my Pro-Q 4 settings, let me know. However, it's not something I intend to use myself. First, it causes unwanted dampening of "sh" and "ch" sounds, necessitating automation to turn it on and off as needed. Second, my settings are not exactly a fine-tuned. It was just something I was messing with in an attempt to fix some harshness in Teto's C5 range. I will not fine-tune it any further, as this noise is less of an issue SVS2, so I'll just wait for Teto 2 to be released rather than fiddle with automation and whatnot.
Note: I still believe that source recording quality can be an issue. For example, the previously mentioned loud plosives found in the original Solaria SVS1 database. (I haven't tried the new version.)
This can be isolated (and removed) using the spectral dynamics feature in Pro-Q 4, but doing so also affects "sh" and "ch" sounds.
Like most unwanted noise, it is much more noticeable on pitched sounds (AKA vowels); and more noticeable on higher notes, and/or vocal databases that have more upper harmonics (AKA audible vowel harmonics that traverse across that 5 to 10 kHz noise zone).
Importantly, this noise is far far far more noticeable on SVS1 voice databases, although it still exists with SVS2 voices.
If anyone's interested in my Pro-Q 4 settings, let me know. However, it's not something I intend to use myself. First, it causes unwanted dampening of "sh" and "ch" sounds, necessitating automation to turn it on and off as needed. Second, my settings are not exactly a fine-tuned. It was just something I was messing with in an attempt to fix some harshness in Teto's C5 range. I will not fine-tune it any further, as this noise is less of an issue SVS2, so I'll just wait for Teto 2 to be released rather than fiddle with automation and whatnot.
Note: I still believe that source recording quality can be an issue. For example, the previously mentioned loud plosives found in the original Solaria SVS1 database. (I haven't tried the new version.)
So, I had an epiphany.
It suddenly occurred to me that if I rendered Teto as two different files using the render "as extra files (isolated)" option, I might get better control over the unwanted engine noise. For those not familiar, the above option generates two files, one containing the "aspiration" components, AKA noise frequencies; and another containing the "periodic" components, AKA pitched frequencies.
Using the render "as extra files" option led me to discover that the unwanted engine noise is only heard in the pitched file. (Or at least that's my current conclusion.) This means that Pro-Q 4's spectral dynamics feature can be used in a far less destructive manner. In practical terms, I used Pro-Q 4 on only the pitched file, then merged it with the untouched noise file. The result was a SynthV Teto that is easier on the ears, yet does not have compromised "sh" and "ch" sounds.
Importantly, this engine noise is found in the same area regardless of the voicebank, which means Pro-Q 4's spectral dynamics are turning out to be very useful. All I have to do now is tweak it to perfection. (Of course, it'll never be perfect, but it helps!)
It suddenly occurred to me that if I rendered Teto as two different files using the render "as extra files (isolated)" option, I might get better control over the unwanted engine noise. For those not familiar, the above option generates two files, one containing the "aspiration" components, AKA noise frequencies; and another containing the "periodic" components, AKA pitched frequencies.
Using the render "as extra files" option led me to discover that the unwanted engine noise is only heard in the pitched file. (Or at least that's my current conclusion.) This means that Pro-Q 4's spectral dynamics feature can be used in a far less destructive manner. In practical terms, I used Pro-Q 4 on only the pitched file, then merged it with the untouched noise file. The result was a SynthV Teto that is easier on the ears, yet does not have compromised "sh" and "ch" sounds.
Importantly, this engine noise is found in the same area regardless of the voicebank, which means Pro-Q 4's spectral dynamics are turning out to be very useful. All I have to do now is tweak it to perfection. (Of course, it'll never be perfect, but it helps!)
I know I couldn't hear it in the one sample I listened to before, but even so I really appreciate this level of dedication and digging into it?? I feel like you should (or could, I'm not telling you what to do) compile all your findings and send them to DT. They ought to hire you as part of their team 
In case anyone wants to test/improve my Pro-Q 4 preset, here it is:
It defaults to -12.0 dB, which is about as low as you would want to go for SVS1 voicebanks. (Well... -11.7 dB is probably closer to the limit that I could hear, but -12.0 dB is a nicer number.)
Maybe -9.0 dB or -6.0 dB should work nicely for SVS2 voicebanks. However, for backup voices I'd go -12.0 dB regardless of voicebank. After all, the engine noise accumulates if you have multiple singers at once, and unwanted effects are less noticeable if you go aggressive on the harmonies rather than the lead.
Remember to only apply it to the pitched file, then merge with the noise file. I personally do this in a separate "preparation" project, then import the result into the project proper.
It defaults to -12.0 dB, which is about as low as you would want to go for SVS1 voicebanks. (Well... -11.7 dB is probably closer to the limit that I could hear, but -12.0 dB is a nicer number.)
Maybe -9.0 dB or -6.0 dB should work nicely for SVS2 voicebanks. However, for backup voices I'd go -12.0 dB regardless of voicebank. After all, the engine noise accumulates if you have multiple singers at once, and unwanted effects are less noticeable if you go aggressive on the harmonies rather than the lead.
Remember to only apply it to the pitched file, then merge with the noise file. I personally do this in a separate "preparation" project, then import the result into the project proper.