UVI Falcon - v4 = 2026 released - rumors, ads, praise, kindergarden, auto-sampling and off-topic inside!

[mdsp]

-may we hope for an optimized CPU behaviour or do you persist on your Argument "good Sound Quality Needs CPU power"...which does not really help. do you intend to work further on that Point?

And to make things clearer, not all IRCAM oscillators are that CPU intensive. The ones that are (IRCAM Stretch and Scrub) are based on IRCAM's enhanced PhaseVocoder called SuperVP. Usually time-stretch / pitch-shift of that quality can't run in real-time but they've heavily optimised it to the point that we can afford to make it real-time. It definitely won't allow a very big polyphony, but that would be a pity not to provide it just because the available polyphony isn't as high as usual.

On top of that, it requires FFT analysis / synthesis which means that the CPU cost isn't uniformly distributed in time (similarly to what happens in an FFT-based convolution reverb) so if you experience dropouts especially at low blockSize, it's usually not because the CPU can't handle it on average, but because the peak CPU costs is bigger than the time available for this particular block.

Finally usually FFT processing also means there's a latency because data has to be buffered before computing the FFT. But because we use it in a MIDI triggered context, latency is not an option, so on the first note trigger we have to process for example 2048 samples even if the current blockSize is only 128, it makes the non-uniform CPU cost more salient.

I hope having such technical details will help people using it wisely, and be a little more indulgent with CPU cost. It's best to know when and how to use it. In many cases using IRCAM granular or the regular Stretch OSC could do the job, but if you need higher quality pitch-shift/time-strech, then there's always a price to pay.
After all we don't expect a convolution reverb to be runned per-voice.

[ThomasHelzle]

Awesome info - I'm taking notes too

Thanks mdsp and otristan.

Cheers,

Tom

[bundoo]

this is a really nice instrument...

i would love a randomise arp feature, something I've been looking for in omni2 as well

fm 8 does it pretty well

[otristan]

this is a really nice instrument...

i would love a randomise arp feature, something I've been looking for in omni2 as well

fm 8 does it pretty well

Randomise what in the arp ? Because there is Random mode in the arp.

[thetechnobear]

I have one small issue, when I press a note, it immediately creates a sound, but it shouldn't since when I call playNote I tell it to use 0 volume

works fine here, with the following script that waits 1s before turning the gain on:

Code: Select all

function onNote(e)
	local id = playNote(e.note, e.velocity ,-1,nil,ch,nil,0)
	wait(1000)
	changeVolume(id, 1)
end

BTW you probably should also have a look at the fade(), fadein(), fadeout() functions.

another question.. which actually I asked before... how can i use

Code: Select all

function sendVoiceModulation	(voiceId,id,targetValue,rampTime )	

We've quickly hinted about it in previous messages but it's not officially available in the engine yet. The idea is to have scriptable polyphonic modulation sources that you can explicitly connect in your patch.

@mdsp
Ok, Ive found the problem....

Code: Select all

changeVolume(id, (e.value / 127.0))

which is equivalent to :

Code: Select all

changeVolume(id, (e.value / 127.0), false,false) 

the issue is the 'immediate' flag...
the above causes a 'spike' .. I'm not quite sure why, since actually the first values I send are very small,
and the spike sounds like it goes to full volume

but changing to

Code: Select all

changeVolume(id, (e.value / 127.0), false,true) 

(so immediate = true) fixes the issue
so I'm wondering is the documentation correct, or the wrong way around?

voiceId id of the voice to be transposed
gain linear gain to be applied between 0 and 1
relative (optional) specify if the transposition is relative to the last changeVolume, default to false
immediate (optional) apply transpostion without smoothing, default to false

its odd since its almost as if the last parameter = true turns on smooth,
OR
it could be just a bug in the smoothing on the initial value

Interestingly I can reproduce the issue easily enough

Code: Select all

function onNote(e)
	local ch = e.channel
	local n = e.note
    local id = playNote(n, e.velocity ,-1,nil,ch,nil,0)
    --wait(1)
    changeVolume(id, 0.1, false,false)
end

play this and you will here the 'blip' at the start of the note
so you can try two things...
- uncomment the wait, and the blip disappears
- alternatively change 'immediate' to true, and it also disappears

for the time being Ive change 'immediate' to true, as i say this works around the issue...
but interested to hear what you find

do we get sendVoiceModulation in the 1.0.1, or need to wait a little longer ... this is going to be so powerful for me. if you need a beta tester for it, drop me a PM, Id be very happy to help as I can test it using my MPE script

[Sampleconstruct]

A little Falcon soundtrack for the evening using three patches and nothing else - multi-sampled extra large frame-drum, granular soundscape, sweeping synth using the pluck synth module.

https://soundcloud.com/sampleconstruct/ ... falcon-air

[wagtunes]

Okay, somebody brought this up and I really didn't think it could be possible that such a basic function couldn't be done so I had to test it out for myself.

A basic function of most analog synths (at least the better ones) is to control the filter cutoff with an envelope so you can get those classic "filter envelope" sounds. It's basic analog synth 101.

Well, much to my surprise, you can't do this with Falcon.

Was this an oversight? Was it done on purpose? I just want to understand the reasoning because I am totally bumfuzzled. Such a basic function not possible in such a complex synth?

Can somebody enlighten me?

[SJ_Digriz]

maybe I don't understand .. but you can add a filter and an envelope to the keygroup and assign the added envelope mod source to the filter cutoff.

[wagtunes]

maybe I don't understand .. but you can add a filter and an envelope to the keygroup and assign the added envelope mod source to the filter cutoff.

Except you can't do this at the program level. Let's say you want ONE filter to control 6 different keygroups. You can't do this. You have to assign a filter to EACH group.

Okay, I guess technically you can still do this, but this should be a basic function that you should be able to execute at the program level, IMO.

[SJ_Digriz]

I'm pretty sure it works at the layer level also but the patch I made like that may be working for a different reason, I haven't tried at the program level.

[IIRs]

Effects at the program or layer level are not polyphonic, hence no traditional envelope control. You need the filter at keygroup level for that.

You can however select a layer or program node, then add a filter to all keygroups in one go, then modulate all those filters with an envelope in one go.

[Sampleconstruct]

maybe I don't understand .. but you can add a filter and an envelope to the keygroup and assign the added envelope mod source to the filter cutoff.

Except you can't do this at the program level. Let's say you want ONE filter to control 6 different keygroups. You can't do this. You have to assign a filter to EACH group.

Okay, I guess technically you can still do this, but this should be a basic function that you should be able to execute at the program level, IMO.

But we have been discussing this (program/layer/keygroup/retrigger) to death many pages back and now you bring it up again, as it was never discussed before. That's some latency here...

[wagtunes]

maybe I don't understand .. but you can add a filter and an envelope to the keygroup and assign the added envelope mod source to the filter cutoff.

Except you can't do this at the program level. Let's say you want ONE filter to control 6 different keygroups. You can't do this. You have to assign a filter to EACH group.

Okay, I guess technically you can still do this, but this should be a basic function that you should be able to execute at the program level, IMO.

But we have been discussing this (program/layer/keygroup/retrigger) to death many pages back and now you bring it up again, as it was never discussed before. That's some latency here...

I guess I missed that part of the discussion Simon. Unfortunately, this thread is getting way too big and I'm having trouble finding things or even whether or not they've been discussed. And sadly, it's only going to get worse as this thread goes.

By the way, thanks for all the demos. They're great.

I only wonder if I've found all of them.

[Sampleconstruct]

maybe I don't understand .. but you can add a filter and an envelope to the keygroup and assign the added envelope mod source to the filter cutoff.

Except you can't do this at the program level. Let's say you want ONE filter to control 6 different keygroups. You can't do this. You have to assign a filter to EACH group.

Okay, I guess technically you can still do this, but this should be a basic function that you should be able to execute at the program level, IMO.

But we have been discussing this (program/layer/keygroup/retrigger) to death many pages back and now you bring it up again, as it was never discussed before. That's some latency here...

I guess I missed that part of the discussion Simon. Unfortunately, this thread is getting way too big and I'm having trouble finding things or even whether or not they've been discussed. And sadly, it's only going to get worse as this thread goes.

By the way, thanks for all the demos. They're great.

I only wonder if I've found all of them.

I thought you were here when we discussed it a few days ago, but what the heck, now you know again

...and thank's, my pleasure, all videos here:


audio:
https://soundcloud.com/sampleconstruct/ ... ibrary-for

[otristan]

1.0.1 is live.
The changelog will be available when the KvR news is posted.

Cheers !