How to fix strange error similar to frqs but not frqs?

My friend has a VB that for some reason, when used on all resamplers other than default and moresampler, sounds like this: https://clyp.it/1wnwxfz5?token=634c1daa581d8eea544fb2482117c246

It’s not frq errors or oto, as they’ve been fixed and oto is moresampler generated but works normal as far as it should in moresampler and default. Frqs in utau look like this:




Resamplers that it has been seen to have this effect on:
- Fresamp
- TIPS
- EFB-GT

Resamplers where it works fine:
- moresampler
- default

It sounds similar to frq errors/modulation/no formant but frqs have been fine and the usts have been cleared, the effect is still the same.

The recordings are all mono and also themselves very clear.
VB itself is not corrupted in any way.
Recorded on a rode ntk
All other VBs recorded in the same way by the same person work fine.

They’re pretty stumped on how to fix it, so if anyone knows what might be the problem, suggestions would be much appreciated

Thank you

Kiyoteru said:

The sample sounds to me like old cached samples are being mixed with the current render, and that the overlaps haven't been crossfaded. However, if it really is resampler-specific, maybe it's possible to edit pmk files/etc. with frqeditor? Since it's able to handle desc.mrq in addition to plain frq files

Click to expand...

They told me they did clear the cache and optimize the ust. I will ask them about the pmk files etc, though
[doublepost=1553815897][/doublepost]

RaccoonButler said:

They told me they did clear the cache and optimize the ust. I will ask them about the pmk files etc, though

Click to expand...

Update: they say no observable change

susrever said:

I know you said that frqs work fine, but did your friend try regenerating them with a different resampler? Because, like, I could understand if it was a problem only with resamplers with proprietary frq formats, but fresamp uses the standard frq files...

IIRC, 0.4+ UTAU versions came with a resampler that sometimes gave problems when generating frequencies (can't remember if it was fixed in later versions, but I definitely remember that when UTAU 0.4 came out lots of people were pointing that out), maybe that's the problem?

I know I'm not much help, but at this point, if nothing else works, I thik trying to regenerating entirely the .fqrs with a different resampler might do the trick. (Also, just curious. What resampler was used to generate the original frqs?)

Click to expand...

Ah, i myself tried that when given the vb, and they also tried putting them through speedwagon i believe- still produced the same results, but thank you still ! They decided ultimately to just rerecord the affected pitch and it seems to be working this time around thankfully