Cannot record pitch automation with MIDI joystick

[gzankevich]

I could no see a way to record pitch automation with a joystick.

I tried:

• selecting all of the possible MIDI inputs in the track configuration then recording
• creating an automation track then recording

System:
LMMS 0.4.15
Arch Linux x64
Korg MicroKontrol

[diizy]

Mislabeled - this is not a bug, it's more of a feature request

[artao]

+25 billion !!!
And not just pitch-wheel and mod-wheel, but ALL the knobs available for automation control. .. And not only controllable via joystick, but via external MIDI device .... Also recordable simple by moving any given knob in the interface with the mouse.

Please!! This really is an essential feature. Having to draw automation envelopes by hand rather destroys any performance spontaneity.

Thanks! =D

[diizy]

It's already on the TODO list.

[tobydox]

We have it in the old master branch already - see

• f40b610
• 198cfe9
• c12b716

If somebody volunteers for porting them to the current master branch, that feature would be available soon!

[diizy]

On 05/04/2014 07:13 PM, Tobias Doerffel wrote:

We have it in the old master branch already - see

• f40b610
  f40b610
• 198cfe9
  198cfe9
• c12b716
  c12b716

If somebody volunteers for porting them to the current master branch,
that feature would be available soon!

I remember trying it in the old master branch but IIRC it wasn't very
functional yet... at least I remember I couldn't get it to work. Might
be some extra work needed to make it work?

Also I didn't much like the way it was implemented in the old master,
there was just checkboxes in the main toolbar for enabling automation
recording, which is IMO a bit troublesome approach... if you enable the
checkbox, will all automation patterns switch to record mode, or how
will it know which ones you want to record... will you then
inadvertently overwrite patterns you don't want to record to... I think
a better approach might be something like what's (although
non-functional) in current sampletracks, where you can switch individual
patterns/tracks to record instead of play...

[diizy]

Also another problem: that implementation seems to be designed to only
work with midi-controllers, so someone without midi-controllers (I
presume that's quite a lot of LMMS users) will have to play with
external virtual MIDI controllers, which is a very clunky solution.

I think a more generic approach built-in to the automationpattern class
itself, which allows recording by either midi controllers OR tweaking
the UI knobs itself, would be much better. Maybe we can somehow add a
switch to AutomationPattern that causes it to instead of "playing back"
automation to instead read and record automation from the first control
its connected to? That way it wouldn't matter what way the automation
would be input to the knob - you could even record LFOs or peak
controllers into automation patterns... much more flexible.

@tobydox, do you think this would be possible to implement? I think it
should be... wherever we have the part that checks for if an
AutomationPattern is muted, just add a check for if it's in a
"recording" state, and then add code that "reverses the dataflow"...
from the control to the automation instead of the other way around.

[diizy]

Sorry for the multiple posts... but I think it might be possible to
adapt the existing code to work in a way I described... I might look
into it at some point (might take a while though as I'll be quite busy
next week).

[tobydox]

Yes I think it would work that way. Basically the arm/unarm-functionality in the mentioned commits at least provides this kind of flag we could use for an improved implementation. We wouldn't need the checkboxes at the top.

[diizy]

On 05/04/2014 08:09 PM, Tobias Doerffel wrote:

Yes I think it would work that way. Basically the
arm/unarm-functionality in the mentioned commits at least provides
this kind of flag we could use for an improved implementation. We
wouldn't need the checkboxes at the top.

@tobydox, I think this might actually be much simpler to implement than
what is done in the commits mentioned... I don't think we need a
separate new "recorder" class to do this.

I'm looking at the code now, and it seems to me that we could simply add
a flag "m_isRecording" (or something) to AutomationPattern, and make
this switchable from the context menu in AutomationPatternView.

Then just modify processMidiTime() in AutomationPattern to check for
m_isRecording, if false, do what it does now (set the value of the
control), if true, do the reverse - read value from control and set the
value in pattern. processMidiTime() is in turn called by the
AutomationTrack's play() function, so this should theoretically work.

I'm going to try this out... there are probably some caveats or problems
I'll run into, but it's at least worth trying...

[diizy]

OMG it works.

It was this simple all along?!

Wow. It really works! I'm currently recording automation by moving the
kicker's vol knob by hand!

[diizy]

Now I'm recording a very neat sinewave pattern from a LFO controller...
this is fun!

[artao]

SWEEEET!

[diizy]

This can be closed now, as automation recording has been implemented.