Acustia Nebula 3

[ronnie]

I've searched around trying to find some detailed info on the Nebula 3 and one thing is eluding me. I've submitted a support ticket to Acustia hoping to get a solution. So far I don't see how it is possible to use Nebula 3 in a live situation as an effect to a Native MIDI VST. There's about a second of latency from the time a note is triggered to the time it exits Nebula 3. I've played with the likely suspects in Nebula 3 based on what I could find out (nothing addresses MIDI latency specifically) with no improvements. The machines I'm using should pose no problems: i7 3.5 Ghz 16 GB RAM Win 8.1 and Quad-Core 2.8 Ghz with 8GB RAM on Win 7. Is Nebula 3 only workable on pure audio for mixing and mastering? I'm not feeling very smart here. I'm more that happy with SCOPE but the buzz around Nebula 3 and Nvidia CUDA has got my ears perked up.

[jhulk]

theres a lot of processing going on and if you read a lot of the plugs for it its better for off line processing when bouncing

[ronnie]

Reply from Acustia Support: "There is not possible with the current audio CORE. We are working in a new technology with less latency but there is not ETA." Does it mean the use of CUDA? I guess using it live via MIDI matters little when you can use a separate Nebula Machine to record and master. I see a problem controlling the mixdown using MIDI automation as the delay would obviously have to be compensated for. When I have more time I'll check it out but I don't see any real "certain something" compelling reason to use it at this point. Perhaps some of the mastering plugs are really as good as people say and those would work in place. SCOPE has me spoiled so that I don't have to bounce with every tweak and just put a controller on anything and change in real time. Ahhh... real-time.

[dawman]

Many guys got Satin 1.2 from u-He and can achieve real time control.
For damn fine sounding ADT effects the latency is really high at 25 msec. but it is a Bidule module customized to trigger the technique and is very impressive.

Sadly it is a Bidule VST dll. inside of the Bidule Standalone, so you would need to demo Bidule, but do not get free demos on their VST plug in version, just the host.

But you can however download Satin and use it as a cumulative effect like Nebula does, and after each track recorded hear it building in intensity for that saturated Bus Compression that's the talk of the town .

[ronnie]

I hear you. Gonna give Presswerk a listen. Also gonna try out this one which looks maybe more promising:

http://www.tokyodawn.net/tdr-kotelnikov/

[kensuguro]

As far as I know, anything convolution seems to introduce a good amount of latency.. what sort of effect in Nebula are you looking to use? There are other convolution plugins like SIR 2 that can do convolution with no latency (aside from ASIO base latency). But SIR 2 is more for reverbs, though technically it's all up to what impulses you feed it. Voxengo's Pristine Space also goes down to pretty low latency. Nebula does a bunch of different things so if you break them apart, there usually are good low latency alternatives that doesn't use convolution to model behavior.

[jksuperstar]

Ken speaks the truth on convolution and latency. It can be calculated as n two ways: in the time domain, which requires as many multiplies and adds per input sample as there are samples in the impulse response, or by multiplying things in the frequency domain, which requires FFT, as many multiplies as your sample rate, and another iFFT. Its either very compute intensive or by using FFT, high latency.

[ronnie]

As far as I know, anything convolution seems to introduce a good amount of latency.. what sort of effect in Nebula are you looking to use? There are other convolution plugins like SIR 2 that can do convolution with no latency (aside from ASIO base latency). But SIR 2 is more for reverbs, though technically it's all up to what impulses you feed it. Voxengo's Pristine Space also goes down to pretty low latency. Nebula does a bunch of different things so if you break them apart, there usually are good low latency alternatives that doesn't use convolution to model behavior.

This is very true and SIR has always been impressive to me. The thing that attracts me about Nebula is of course the sound quality but also the concept behind it: A malleable architecture as an open platform for a variety of acoustic processing with a comprehensive base of adjustable parameters with offline "real-time" processing via CUDA. Sounds a lot like the Open Scope vision to me. I guess the basic assumption of the developers of Nebula is that the PC along with CUDA has come of age where native devices can have more breathing room. They are not there yet but still are usable with some effort. The sound quality is there if you are willing to tweak to it to your system and render along the way (which I am not). Hopefully if and when they get the CUDA routines working this will improve significantly.

[dawman]

CUDA on an old ATI Graphics Card back in the late 90s using a Reverb was really cool, beat the pants off of the metallic reverbs Native had then, I was on the BETA team and one day NVidia came and bought them out, no mas CUDA....

Needless to say using the GPU for FX was a great idea as Gigastudio had no Reverb at that time either, we were upset.
NVidia didn't do jack shit with it for years yet had a strategic advantage with it.

Would love to see the CUDA stuff as I have the HD4600 Intel GFX on my CPU and it barely gets used as our 2D apps need very little.

Actually invented by nerds @ UBC in the Bay Area, I bet they got some nice grants after that success story.

[ronnie]

It seems to me that the CUDA coding tools are out there for the taking but only the CAD and scientific communities have jumped aboard. I agree that this could be a great platform for native plugins. I could see a cross-compiler for Scope to CUDA and vice versa.

[astroman]

forget about it - that's 2 completely different worlds...
(not even considering the fact that Scope is a compiler on it's own)

CUDA is for massive parallel processing of the most stupid kind of calculations
that's what geometry is about and that's where it shines...
DSP (in musical context) is right the opposite: a sequential process of complex (often non-linear) rules

the only strategy where the 2 meet is (you guessed it...) convolution
so it's quite a natural choice that Nebula supports CUDA
and they probably have an advantage over the static approach with it's single sample response
(as explained on their website)

on one hand I don't trust in the method as it's just about different levels
for obvious reason data isn't continous in the time domain and it's hard to imagine complex interactions
(when circuits are momentarily overloaded in real life)
but I suspect those dudes from Positive Grid (with the iPad app BIAS I frequently mention) just do something alike

I'm pretty shure that convolution is a part of the processing in some way (as in many amp simulations)
but the tone is less static than expected
considering a pure algorithmic approach the result is way too variable
it lacks the common siganture all(!) the other plugins produce in some way
as a third hint the output of BIAS is significantly less crisp and defined in comparison to DSP processing
... but then it sounds so much more 'realistic' and kind of pleasing to the ear... leaves me clueless
(btw in a recent article about virtual guitar amps it was called the best sounding modeller available)

I wouldn't be surprised if something like Nebula's strategy was applied in a modified context

cheers, Tom