MSF: Burn transformed oscillator cycle to wav and load into oscillator section sample loader to save CPU?

[Annabanna]

I just had a thought, and was wondering if this is currently doable in a reasonably easy fashion, or whether it's the type of feature that might be worth adding down the line.

After spending quite a bit of time creating custom algorithms for mTurboReverb, then burning them to IR's to use in mConvolution, by sessions are running SOOOOO much smoother without all those big CPU hits. Frankly, unless I'm doing something that CAN'T be captured by IR for some reason or suddenly have unlimited cpu cycles, I don't see myself working with reverb any other way from now on.

Similarly, I was reading up in the manual for MSF about the CPU usage for waveform creation, and noticing the significance of the various quality settings and their CPU consumption, and it dawned on me as yet another instrument I was creating crippled my system that this might be partially solved in a similar way.

If I have already used the tools in the oscillator section to shape my waveform exactly how I want, then why am I burning all that CPU power from that point forward? If it's just a static waveform, and I'm not modulating those parameters, then wouldn't it make sense to be able to sample that customized waveform at it's highest possible cpu crushing setting, load that into the sample loader in the oscillator section, and save my cpu for other tasks?

Is anyone doing something along these lines already? Reasons why it makes sense or not to be doing?

[Lairyboy]

It sounds like a reasonable idea if you take multiple samples or play within a narrow frequency range. You lose tweakability and might miss some subtle timbres, or indeed introduce them by stretching your sample vs straight from the oscillator.

[Annabanna]

... and might miss some subtle timbres, or indeed introduce them by stretching your sample vs straight from the oscillator.

Yeah, that's the part I'm less sure about. It's understood that this wouldn't work if you're animating oscillator properties just like it doesnt' work to to the IR clone of mTurboReverb if you've got modulation baked into the algorithm.

I have no idea, though, how problematic it is that the sample in this case would be of a wave cycle at a fixed sample rate which may then be stretched. In the case I've been working on for the past week, it's all basically customized 808's, so the range is int he worst case still under two octaves.

What makes me think it's doable is Omnisphere. Maybe I'm wrong about this, but I was under the impression that huge bank they have of waveforms are just stright up samples. I mainly use it for bass sounds, so the range isn't huge, but I've never noticed any issues in Omnisphere using the samples vs oscillators.

The goal here too, though, is to dial in an insane level of accuracy that would just be undoable otherwise... like super crisp instant attacks, etc that get smeared with lesser quality settings. I did the same thing in mTurboReverb. The IR's I'm using now in some cases have 45 parallel instances of the most complex algorithms running at maxed out complexity. It takes MINUTES to burn a single IR, so obviously it's not even doable any other way as it can't be computed live. i had to develop a special A/B comparison device with a dumbed down test patch of each algorithm just to estimate what the results would be. Similarly, I want to burn waveforms at the highest possible setting that would crash the session live.

I'm interested to try it. What would bre the best way in Logic to quickly capture a single waveform cycle that wouldn't require any trimming or offset issues from latency, etc?

[Lairyboy]

I'm not a logic user, I'm afraid so I wouldn't know where to start for capturing a single cycle as described. Would you even want to catch just one cycle? Wouldn't there be variation over cycles that would give your sound it's own timbre. Taking several samples over intervals and using multisamples might give the best fidelity.

[HorusAnd]

My understanding is that, transformation modulation is the oscillators biggest CPU hog. But that's kind of irrelevant here. If you bounce the audio and reimport it into the oscillator, the oscillator import function analyzes the waveform and then uses the oscillator to replicate it as close as possible (with mixed results, it wont sound exactly like your original waveform and the modulation doesn't really make it through the analysis like you might think). Nonetheless, you're still using the oscillator which is going to use about the same amount of CPU barring whatever modulation was being used on the original oscillator.

You mentioned Omnisphere. Omnisphere is basically importing the waveform into a sampler and playing back a waveform with loop points. < This method will drastically reduce your CPU as well as retain a, more or less, perfect copy of your oscillator, barring inconsistencies that may occur over long periods of time from parameters like the "analog drift".

I use the latter method all the time because MSF cripples my CPU with more than a couple instances. Even my M1 mac can't really handle it lol

Algorithmic Reverb = IR Convolution
Synthesis = Sampler

Hope that helps

[Annabanna]

I use the latter method all the time...

Meaning you burn the oscillator waveform, and load it into a sampler module within MSF? (Haven't checked loop options there as I mainly use sampler for drum one-shots).

Do you have some internal way of doing that, or are you bouncing through your DAW? Any snags there, or tips/tricks to avoid misalignment? (I seem to be having a number of issues with Logic bounces lately not lining up as well as they should. Not sure if it's a latency issue or what yet)

[HorusAnd]

I use the latter method all the time...

Meaning you burn the oscillator waveform, and load it into a sampler module within MSF? (Haven't checked loop options there as I mainly use sampler for drum one-shots).

Do you have some internal way of doing that, or are you bouncing through your DAW? Any snags there, or tips/tricks to avoid misalignment? (I seem to be having a number of issues with Logic bounces lately not lining up as well as they should. Not sure if it's a latency issue or what yet)

Yup, I bounce (burn) usually C3, (depending on the sound, the length varies), using my DAW and then I drag that file back into MSF.

Some sort of internal way to do it would be awesome! But, afaik there isn't.

Any alignment issues should be fixable within the sampler by adjusting the sample start. Unless Logic is somehow chopping off the beginning when you're bouncing?

You might get a better result by importing it as a wavetable in 1024 and 2048 cycle sizes. But there's maths involved that I don't really understand. But that might be a much better solution as you can control it using the "Wave" (index).