Convolver woes when rendering at higher sample rate

I’ve got songs I’ve been working on at 48kHz with several convolver instances running 48k and 44.1k impulse responses, with everything sounding as intended. When I render at 96kHz or 192kHz, however, the first 2 seconds or so of the output seems to have the convolver output smashed, until it clicks into where it should be, and from that point on the output is as it should be. It’s like the rendering starts before the renderer figures out how it should resample the convolver output, and only after does it render correctly.

Here’s an xrns with example output:
convolver-render-issue.xrns (2.6 MB)

I’m on a laptop most of the time (and thus limited to 48k for output/realtime rendering), so having to fire up my studio computer to run through my interface at 96k or 192k just to render a track in high quality is a bit soul-crushing.

I’ve tried all manner of render settings, sample interpolation/oversampling, etc – to no avail. Best I can tell, the problem is with the convolver’s resampling, but I don’t want to send anyone down the wrong rabbit hole.

Help? :cry:

Can’t help, but it sounds a bit like this issue: First note missing from Render

Why not just use 48 kHz, it should be completely sufficient.

1 Like

It may be related, but I’m not using any plugins, just pure samples, native FX, and impulse responses.

It’s not. :confused:

There are Gremlins in the render function, Taktik should take a close look at what’s going on there.

Does it happen, if you replay the song in 96kHz, too? If not, maybe try to use melda mrecorder instead

If I set the Audio output settings to 96kHz (through the interface), and I render at 96kHz (realtime or otherwise), it works as expected—my impulse responses keep their length and character, and no audio artifacts appear. It’s only if my interface is set to something else that it doesn’t render right. (i.e., If I’m on my laptop with 48k built-in interface, I can’t render to anything above 48kHz without convolver artifacts).

This is curious to me, though, since I would expect the interface “live” sample rate to be completely irrelevant to the render process. To me, anyway, an ideal renderer should be able to output viable audio files even on a computer without any audio interface whatsoever.

If it comes to it, I’ll give mrecorder a shot…

That is interesting. Or not.