Here are links to the Mac beta versions of everything. These should fix the major outstanding bugs and improve performance significantly. Fixes include:
- fixed dial display updates when changing number of voices
- fixed hanging notes problem with some hosts including FL Studio, Reaper
- fixed an issue with dial quantizing
- fixed a problem that could cause Kaivo audio to freeze
CPU use for Aalto and Kaivo should be reduced by around 20%.
Please see the readme files in the installers for more info.
Windows versions and a release will be coming early next week, hopefully.
http://madronalabs.com/media/aalto/Aalto1.8.2b3.pkg
just tried these with a test project with all 3 plugs - current versions ran at 70%, beta versions ran at 50%. great work, thanks randy!
to follow up, lots of my patches are sounding messed up with Aalto now - email sent thanks.
Noticed that setting rate to 0.0 doesn't seem to work as expected. No grains are triggered in the "arf!" patch, for example.
I'm making progress here. To get some of these big speedups I moved the plugins onto my new madronalib framework sooner than I had planned to. (More on that soon.)
Unfortunately there was a big hurdle with this new and awesome code on the 32-bit Windows platform in particular. It's taken me some time to figure out a solution. I just compiled something that works on Win32, so I'm hopeful that I can get to these other reported issues and release soon.
Clearly there's been a bit of work in the area of the sequencer, quantisation and range .. no doubt this is a nuanced issue, however, this strikes me as wrong or significantly counterintuitive
Currently if you start from scratch and only connect the seq output to pitch then there's a significant change (explore by dragging the slider fully up/down)
Previously +12 would give you an octave range, now to get an octave range you have to add 11 (or -11)
Quantization is working at the extremes, but is creating problems in smaller ranges due to the algorithm used
Try +1 and note that quantisation makes no difference, it doesn't quantise
Try -1 and you see that quantisation is immediate as you rise from zero
Try +6q and +7q ... they both behave the same, they both have only 6 additional pitches
Clearly there's a +1 or -1 issue in an algorithm somewhere because by the time it gets to what should be an octave it is gone askew .. well, maybe not so clearly, but something is amiss partly with the quantise which may be a tricky area to get right but also to the range
Maybe this graphic can help explain, in my opinion it ought to be possible to have a consistent range and predictable quantise points so it works well for quantise on/off
You can see in this example that to hear an octave range you need to hear 13 distinct frequencies, but to achieve an octave in the beta, you only hear 12 distinct frequencies in total, so the quantized steps cannot be in semitones in the beta
I hope this sheds some light on it, i’m glad there’s been some movement in this area (yay, at last), but there’s clearly an issue or two at play
^ you'll need to open that image in a new tab, clearly scaled too large by default
maybe randy can scale it down for a preview inline
Thanks for the comments. I could not reproduce the issue with 0 giving any change or +12 not being an octave.
Are you sure you have your attenuverter set to 1.0? Try double-clicking it to restore the default.
Do you have glide on? If so, the output value will always be a linear combination of two adjoining sliders based on the current time value. So if you are doing testing like this you have to make sure glide is off, even when the rate is 0.
I tried all the quantize ranges, and I did find some oddness with |range|
Well, that's a bit awkward, i've clearly managed to nudge the attenuverter .. it's a common issue with the way i hold the magic mouse with fingers hovering above (and too close) to the sensitive virtual scroll ! That accounts for the octave coming in at 11, so that aspect can be put to bed. Glide was off fwiw.
A quick look at the situation as it is puts it back closer to the way it was before (whilst mousing a slider now seems to take it to its Max i.e. no need for scrolling the last %)
So, fwiw, in terms of the friendliness of the quantising from an end user's perspective i noted the following
with a Range covering the first octave each way in semitones
those with the symbol are all quantised down a semitone
+12▼ +11 +10▼ +9▼ +8 +7▼
+6 +5 +4▼ +3 +2 +1▼
-1 -2 -3 -4 -5 -6
-7 -8 -9 -10 -11 -12
same with +24,+36 and so on all ▼
This was one of the issues i raised before about 'expectation' from the quantisation, so without knowing if the strategy i suggested has its own issues, it would at least circumvent these issues
The current quantisation isn't predictable imho, based on the observation above
One other comment, i'm curious, has anything changed with regards to the output of the sound engine in a manner which might affect the tone, something isn't quite right with my presets, i can't quite put my finger on it .. a little bit tamer, i looked through quite a few of my presets and was wondering if the timbre settings had been rolled back somewhat
I should have pointed out that the above was for Aalto AU in Max on Mac fwiw
Re Kaivo
A lot of presets i had which were functional in the last release are no longer functional . these exploited the previous behaviour whilst rate was 0.00
I also note that when changing between presets which both share a Rate of 1.0, the rate dial will momentarily jump up to 220 for a split second then revert to 1.0, it is interrupting audio
Has something fundamental happened in this area to result in the change in behaviour between the last release and this beta, do these observations make sense ?
I think i'll revisit the last release to see why i'd programmed them that way
One other comment, i'm curious, has anything changed with regards to the output of the sound engine in a manner which might affect the tone, something isn't quite right with my presets, i can't quite put my finger on it ..
It's possible that the sound has changed very subtly. My intention was not to change the sound, however many things have been rewritten. When I make changes to the audio engine I listen to a series of reference patches I know very well. If these sound as they should, then all is well.
If you have a patch that sounds different enough between versions as to constitute a bug, you are welcome to send it to me for listening.
A lot of presets i had which were functional in the last release are no longer functional . these exploited the previous behaviour whilst rate was 0.00
This was pointed out here and fixed.
I also note that when changing between presets which both share a Rate of 1.0, the rate dial will momentarily jump up to 220 for a split second then revert to 1.0, it is interrupting audio
That's new to me, I'll investigate.
OK, these should be the final betas. I'm planning to release on Monday.
I fixed an issue with the mono->stereo processing and channels for Virta. Hopefully this works better for people who were having trouble using mono synths.
Mac:
http://madronalabs.com/media/aalto/Aalto1.8.2b5.pkg
http://madronalabs.com/media/kaivo/Kaivo1.3.2b5.pkg
http://madronalabs.com/media/virta/Virta1.3.2b5.pkg
Win:
http://madronalabs.com/media/aalto/AaltoInstaller1.8.2b5.exe
http://madronalabs.com/media/kaivo/KaivoInstaller1.3.2b5.exe
http://madronalabs.com/media/virta/VirtaInstaller1.3.2b5.exe
Does this previous post about the mono->stereo fix the issue with Logic X not outputting stereo? If so, thanks so much!
One hopes, please try it.
With granulator Rate set to zero it was possible to play through the sample linearly (and loop it)
It is not possible to modulate any of the rate dials when they are set to zero, so i don't see the harm in preserving that option as a special use case .. it adds functionality
I can see no way to play through the sample linearly now, i made quite a few patches using that. Could that linear playback be implemented by another means ?
Also any thoughts on the quantisation issue posted above ?
EDIT: Just tried latest beta and the Linear Playback is back - great !
Feeding the granulator with a Trig allows you to do other stuff to the sample and to feed it through the resonator
Randy: Just tried in Logix X 10.3.1, still left output only if you add Virta on a standalone software instrument track. I will test with a mono input source.
Here's how I tested your issue:
- open Logic
- add a new mono track with an "ES M" (track will be mono)
- add Virta in the Audio FX slot (track will turn to stereo)
now Virta takes the mono input and generates stereo output. It also works with a stereo synth as before.
Can you confirm you see version 1.3.2b5 at upper right? If so please give me detailed info of how I can reproduce your issue.
That works. Which was the major issue.
If I open logic
Add software instrument track
Select virta stereo as software instrument
Only the left channel is output
Admittedly this is not a big deal because I don't use it like that very much. Thanks for making mono inputs stereo outputs!
I see! I can reproduce this. I never tried it before because I never load Virta as an instrument. Out of all the DAWs I have, it looks like Logic alone presents Virta as a software instrument. I'm not sure why it does this. Virta is intended to be used as a MIDI-controlled audio effect. I'll see if I can stop it from showing up as a software instrument in Logic.
Also any thoughts on the quantisation issue posted above ?
It's clearly a bit out of whack, and something I will look at after this version is out. I've noticed that certain scales are not loading correctly, so a review of all scales / tunings / pitch related code is probably in order.
Virta 1.3.2 in Logic 10.3.1 (OSX 10.12.3)
Virta now presents as both an audio effect and Software Instrument (AU Midi Controlled Effect).
Both seem to work so I guess this is a good thing.