i think i read somwhere that,
certain synth(like pro52 and B4) use sampled
waveforms and then process them into a filter that is hardwired on the intrument.

it would therefore make sense that pro-52 and B4 can be used as FX plugins, since its just a matter of directing the audio flow thru that filter or leslie/drive

But there are some synth that generate in-realtime Waveforms. muon's familly and TC's mercury come to mind.

since those two generate in realtime, there is virtualy no aliasing if done right...as opposed to synths that used sampled waveforms(pro-52 and B4)

But the drawback is that they consume more
Cpu (but its worth it!)

getting to the point: what would really be cool would be
the realtime-synth would have a little feature,let's call it generate offline.

when you do you patch and your happy with it, and everything is honky dorry and its perfect and doesn't need further fiddling , you would press "generate offline" and it would generate the patch as samples.

And now it would be like a semi-sampler
and would therefore not need as much Cpu and
would be of better quality than the synths that just send sampled waveforms thru a filter pseudo-plugin.

does that make sense?

ooooooh, wish I'd thought of that.....

Good idea, but you might end up filling up a hard drive with generated wavs.....

Someone please explain me what 'aliasing' is in sound generating/processing?

Thanks.

Hmmmmm. Strange question.

Alot of synths can already do this if you think about it. I mean a few synths allow you to have user wave forms as an oscillator. You could render the synths output and then load it back up into the user wave slot. I imagine Cronox is really powerful in this area and Pentagon I would also be good at this. However, to me it's like loading a single wav sample into a freeware tracker program or simple sampler.

If you have FruityLoops you can just whack the output into a sampler channel and go for it using much less CPU.

However, the reason I don't like many of these approaches is that if you are using a single sample you are going to be severely limited by that actual sample - going more than a few notes up or down the scale from the root note will make it sound quite different from what would be achieved by the real time wav generation of a softsynth.

This is not necessarily a bad thing if this is what you want. And in fact sometimes this would sound better than the more CPU intensive option if you ended up finding the sound you were looking for.

Caleb

I'd like to reply to Caleb:

I've had good results with the Cronox with just one sample, granted there are limits, mostly on the top end, but the results have been better than I had hoped for. It's an amazing little synth.

I'll have s couple of free patches on my website soon. Love to hear your feedback on them.

Kenfen from Kenfen.com

> Someone please explain me what 'aliasing' is in
> sound generating/processing?

Aliasing happens because digital audio can only reproduce frequencies up to half the sample rate. A simple synth oscillator generates harmonics at higher frequencies than this, which end up being "aliased" down to below half the sample rate, where they come out as musically unrelated frequencies.

If you open up the Pro52 filter and select a single sawtooth wave you might hear that some high notes sound "rougher" than others - this is aliasing noise.

Almost *all* synths use some sort of anti-aliasing as without it they would sound awful. The difference is how much CPU the designers decided to spend on it - so Pro52 has quite basic anti-aliasing (it seems to be linear interpolation) but the TC Mercury has excellent anti-aliasing (like a good hardware synth) but with CPU use to match.

Most developers (including me) are now using something between these two extremes, to get the best balance between quality and CPU use, and as processors get faster the balance will change.

Sample playback also requires anti-aliasing if the sample is being played at different pitches, but it doesn't need to be as good as the anti-aliasing used for synth waveforms which usually have a lot more high frequency content.

[ 16 October 2001: Message edited by: mda ]

mda : thanks for the information, I didn't know that.

Now that I do know it, how feasible is it to have switchable or selectable aliasing in a VSTi synth? Ie. I could select whether I wanted none or high-quality aliasing or maybe select from a range inbetween. Sure it might sound nasty but I'd be into that

Or do some VSTi already have this?

Sorry to butt in but are you the same CiM who released Reference on deFocus? Great record.

@CiM: QuadraSID has a selectable "oversampling" per chip which helps a lot with aliasing. You can choose between 1x (no oversampling, much aliasing) to 8x (eight times oversampling, a lot less aliasing). So you can decide which sound you prefer or if your bass-sounds really needs it at all...

I reckon thatz not a bad idea...being able to select the kind/quality of anti-aliasing on a vsti. Alot of tekno is based on nasty noize...the nastier the better for some, so it mite be nice to able to choose (+ compare). Stray crap can sometimes do wonders with a bit of filterage and reverbitude. 8¬)

quote:

---

Originally posted by Michael Kleps from reFX:
It's much harder to generate alias-free waveforms in realtime than playing back samples.

---

Well then if its sample play back it would be less demanding, now wouldn't it?

maybe i didn't express myself coherently.

I'd like to have a VSTi that when I finished doing the patch (filter, enveloppe, transpos etc),
it would generate the samples, and then the synth would playback the exact audio output it would if i was playing it while it outputing in realtime. therefore it would be not so cpu intense.

And another use it would be excellent for, would to make synth patches for Romplers ala Sampltank...and to an extent
Halion(the filter would not be used cause the samples are that of the settings of the above mentionned synth...no need for those crappy halion filter)

did that make anymore Sense

realmarco: Pro-52 is also generating it's waveforms in realtime. It's much harder to generate alias-free waveforms in realtime than playing back samples. The sound-scaping features of a synth (filters, effects etc.) are taking 90% of the CPU and the oscillators only the remaining 10%. Sadly there would be no gain in the route you mentioned. Bypassing the filters and effects would give you a very simple sample-player, and for that you can use HALion, VSampler etc.

Sorry

Glitch: yes and thanks

Michael reFX : great. I think that could be one interesting area that could give a VSTi its own sound - where the methods developers use to get the sound to be CPU efficient or to sound more accurate (or whatever) could in some way be tweaked themselves - like aliasing. Perhaps directly tweaking some of the algorithms in some way. Obviously this stuff is unnecessary in emulating already existing synths (subtractive, FM, etc) - but going beyond that and creating new synths it might be interesting...

I presume things like aliasing are preset in programs like Reaktor or SynC yes? I would've thought thats the kind of stuff in the back-end of Reaktor which remains uneditable to the users.

It should be obvious by now that I don't know what I'm talking about - just some idle thoughts

quote:

---

Originally posted by realmarco:
I'd like to have a VSTi that when I finished doing the patch (filter, enveloppe, transpos etc),
it would generate the samples, and then the synth would playback the exact audio output it would if i was playing it while it outputing in realtime. therefore it would be not so cpu intense.

---

Why not export the track then? It's much easier on CPU (you can switch the VSTi off) and you don't lose anything. When doing what you suggest, the synthesizer would need to generate a multitude of samples (per note, per velocity, per LFO, per envelope etc.) and that is simply not practical in terms of memory. You would also have no advantage above a generated audio-track and that can be done with Cubase etc. in a matter of seconds with every VSTi without changing one line of code.

Sorry, I simply don't see the point. Perhaps it's all a big misunderstanding. Are you speaking german? If yes, we could continue this in private by telephone?

i agree with what most people said here, it doesnt really make sense. the moment you use a modulation routing like velocity-to-filter, you cant just resample a synth sound like that. thats the problem most 80s and early 90s romplers struggled with :-)

but if you care for a way thats pretty fast and comes very very close to your suggestion, here it is:

1. as preparation, make yourself a nice "render VSTI" project in cubase.
it should have 88 midi tracks, each track playing one key. also, even if it means a little typing work, name each track, "01-a0", "02-a#0", "03-b0" etc. open some VSTI, and set all tracks output to that VSTI.

2. now that youve done that, when you have a patch in a song that you want to "render": save the patch to disk and open your "render VSTI" project

3. exchange the VSTI with the one you want to render. all 88 tracks output will switch to the new VSTI, saving you a lot of clicking

4. load your patch from disk

5. press control+A (select all)

6. in the master section, press export audio, in the following settings dialog, select "bounce selection" and not "bounce within locators" (not sure if thats what it says there, my VST speaks german :-)

7. VST will now render all tracks individually, in one giant batch process so to say, resulting in 88 wav files, named "01-a0" all the way to "88-c8".

8. almost all VSTI sampling plugins have the option these days, that you can load a lot of wav files and they will be mapped chromatically, automatically. meaning, you can make your perfect sampler patch with two mouse clicks. well, at least vsampler does this, i assume the others do too (never tried with halion, i use that as preset player mostly)

...there you have it :-)

personally i really doubt this makes much sense, but if you want to try... and its really a very simple thing, im just not very good at explaining it fwiw, i resampled the whole m-tron plugin within half an hour this way, includding mapping the resulting samples and all. for that it does make sense, cause it allows you to play those nice m-tron samples in your sampling plugin of choice, giving you multitimbrality, filtering, panning and all this stuff the mtron plugin doesnt have.

btw, vsampler also got an option to choose between different interpolation methods. so does halion, allthough with halion im not sure if its about interpolation, but its got that nasty quality-vs-performance slider fwiw

derek thats exactly what I wanted
to say (english is not my mother tongue....
its french..ouch hey stop pelting fruits at me!)

What i wanted is a more easier way to make synth samples in a sampler...Do you all remember Simsynth?

kinda like that but being able to load them (samples) into a sampler,or the vsti in question could b the host of those samples.

Thus less cpu , right?

quote:

---

Originally posted by Michael Kleps from reFX:


Why not export the track then? It's much easier on CPU (you can switch the VSTi off) and you don't lose anything. When doing what you suggest, the synthesizer would need to generate a multitude of samples (per note, per velocity, per LFO, per envelope etc.) and that is simply not practical in terms of memory. You would also have no advantage above a generated audio-track and that can be done with Cubase etc. in a matter of seconds with every VSTi without changing one line of code.

---

That's put it much better than I did Michael.

I'm guessing the disadvantage to that method is that you may not have actually composed the music yet. In which case what could you render but a single note?

So I can understand where the original idea is coming from but you would have to render a fairly long note to maintain the complexity of modulations programmed into the sound and even then it could be almost impossible to faithfully reproduce. And furthermore, once you started playing a melody, the sound would change character the further it moved away from the root note and the modulations recorded would speed up or slow down.

Maybe you don't find that imperfection terribly important during the composition phase. However, since you can download a freeware sampler that will quite happily pop a one-shot sample into it and map it across the keyboard (3S is one that comes to mind), there wouldn't be any need for a developer to convert a VSTi into a quasi sampler at the push of a button.

And kenfen - I probably didn't explain myself that well. I was referring to Cronox as a sample player rather than a sound design tool. I'm sure it's capable of producing fantastic sounds using one-shot samples. It's probably not as great at faithfully reproducing the playability of another synth simply by loading a one shot sample of that synth and playing up and down the keyboard.

Caleb

sure, if you can live with the limitations (and have a lot of RAM ;-)

try it, its really easy and fast. it takes a while to create that rendering project template, but once you have done that, rendering any plugin output key by key is done within minutes

Derek, cool idea. Would you mind emailing me your "render" VST Song?
I know - I'm such a lazybone.

Robert.
bluebus@gmx.net

Me too please.

I'll be happier than a pig in shit I promise.