Trouble dialing in precise recording delay compensation in simple case with no plugins

Cubase
1

Cubase Studio 14, Mac m4 mini, Apollo Twin x, SoundSource

Day 1 of trying to return to Cubase after decades on other DAWs, and it’s not going well.

I’m only using Cubase for recording vocals so I can use Melodyne with ARA afterward since my main DAW doesn’t have ARA. I’ve only got 2 tracks, a fully mixed karaoke track and a guide vocal. Both are output from Bitwig, and both are sample-accurate in their timing (beat hits EXACTLY on beat).

I tried recording vocals in cubase, and got horrific delay, so I tried measuring it. I ran headphone straight back into Mic and I’m getting something like 72ms of delay. I dug through all the settings I could find, but no buffer setting, etc even comes close to fixing this. It is clearly not compensating correctly for either the Apollo Twin x or SoundSource that’s running my system audio. In either case, what I need AFAIK is to input a specific offset to bring it back into alignment. In theory this should be straight-forward since I have zero plugins or other variables which would cause the latency to change.

Unfortunately, I’m not seeing a way to do that. The Studio Setup does not appear to allow me to alter the Input or Output values directly, the input delay compensation in the Open Config App popup under “Hardware” doesn’t allow me to input a specific value (and none of the presets even come close).

I’ve tried all the usual garbage like changing the buffer size, constain delay compensation, etc. None of that even comes close, but that’s also missing the point that I’m not trying to minimize the delay here, but rather to measure it an eliminate it entirely by offsetting it by a specific amount.

Have I missed some way to force it to correctly compensate for the hardware latency automatically?
If not, then have I missed some way to take the measured value of the mismatch and input that exact amount to offset future recordings automatically?

Again, this should be a very simple case since I will only ever have two audio tracks (plus whatever I am recording), and zero fx. The goal is sample-accurate alignment, so if I record the headphone directly into the mic, the measured waveform should be less than 1 sample off from the waveform of the guide track.

2

Studio Setup/Audio System/ Record shift. (Make sure adjust for record latency is ticked of course)…at least this is where you set the offset in Windows.

I would guess that Soundsource is what’s throwing things off by such a large amount.
Can’t you do everything you need with the Apollo as a single device.

My offset (for incorrect interface reporting) is a more typical 70 samples…72 ms is huge.

You should measure with a cable loopback for most accurate result.

3

I’ll look up the Record Shift. I don’t remember seeing it.

Yeah, it seems unreasonably high. I’ll look into bypassing SoundSource to test.

As for the cable, I think I have a different philosophy on that. I’m not interested in the theoretical latency, but rather the exact latency as it occurs while singing, and while singing, I’ll be listening to the headphones, so perfect alignment would be that which is perfectly aligned to what I’m hearing as I sing.

4

Thanks for the info Grim.

5

What are you try to say? I don’t get it.

The apollo has a hardware mixer and would introduce very little latency to the monitor path while recording.
It reports that value correct to the DAW and the vocal recording is aligned with the playback.

6

Indeed. The variable here is the use of “SoundSource,” which is a cool app, but shouldn’t be involved anywhere in this equation. Dude said he was going to bypass it and didn’t reply, so I figured that was it. With the Twin he can most definitely record with “zero” latency while monitoring. You can even set up monitoring via cues and swap back and forth between software monitoring and cue monitoring, etc. My guess is SoundSource was mucking up the works :slight_smile: