v6.70+dev1117 - November 17 2022

[tonalstates]

v6.70+dev1117 - November 17 2022

• * Includes feature branch: razor edits on master track envelopes
• * Includes feature branch: selected media item appearance changes
• * Includes feature branch: track grouping manager dialog
• * Includes feature branch: CLAP plugin support
• * Includes feature branch: track media/razor edit grouping
• * Includes feature branch: improve experimental silent-track CPU reduction option to include FX tail length
• * Includes feature branch: media item fixed lanes
• * Includes feature branch: internal pin management overhaul for future extension
• + JSFX: support :sqr fader scaling, default is square, can override power with :sqr=3 e.g.
• + Localization: support UTF-8 in time-signature marker editor combo box
• + MIDI: fix recording of multiple takes via retroactive record [t=272770]
• + Media Explorer: improve search performance with very large databases [t=270722]
• + Media explorer: support zoom/scroll via mousewheel or horizontal mousewheel
• + ReaVerb: fix deconvolution tail trimming accuracy
• + ReaVerb: support deconvolution of more than 2 channels [t=272747] [p=2600567]
• + Render: do not automatically set render source to render matrix when opening region manager [t=272752]
• + Web interface: fix fancier.html record armed track detection with a single track [t=272428]
• # JSFX: fix display of file-based sliders in automation etc
• # JSFX: internal file-slider overhaul
• # JSFX: log-scaling affects envelopes/knobs (use :log! for log UI only and to export parameters linearly)

This thread is for pre-release features discussion. Use the Feature Requests forum for other requests.

Changelog - Pre-Releases

Generated by X-Raym's REAPER ChangeLog to BBCode

[mawi]

+ JSFX: support :sqr fader scaling, default is square, can override power with :sqr=3 e.g.

Works, thanks!

# JSFX: log-scaling affects envelopes/knobs (use :log! for log UI only and to export parameters linearly)

Very nice, thanks a lot!

It would be nice if the ReaPlugins with nonlinear slider scaling would also have their envelopes/param-knobs adjusted (e.g. ReaComp Ratio).

[ovnis]

[*]+ Media explorer: support zoom/scroll via mousewheel or horizontal mousewheel

Is it about the waveform?!

[schwa]

btw this could/should be a significant improvement (factor of 10 speedup in our tests with a database containing ~1 million files with metadata).

+ Media Explorer: improve search performance with very large databases [t=270722]

[tonalstates]

Originally Posted by schwa

btw this could/should be a significant improvement (factor of 10 speedup in our tests with a database containing ~1 million files with metadata).

Just tried it and wow, snappy! and I'm getting more results than before, thank you guys for this.

[tohubohu]

Originally Posted by tonalstates

v6.70+dev1117 - November 17 2022[*]+ Localization: support UTF-8 in time-signature marker editor combo box

OK!
there are still a few other things to fix...https://forum.cockos.com/showthread.php?t=271122
and then an important future issue number 14...after that it will probably be over for the langpack. thanks

[vitalker]

Originally Posted by tonalstates

[*]+ Localization: support UTF-8 in time-signature marker editor combo box

Can't UTF-8 be supported everywhere, so such things won't happen anymore? Or is it difficult? I saw you guys have added UTF-8 support for many things, but still there are places where only ASCII is supported. Why is it so?

[Justin]

Originally Posted by vitalker

Can't UTF-8 be supported everywhere, so such things won't happen anymore? Or is it difficult? I saw you guys have added UTF-8 support for many things, but still there are places where only ASCII is supported. Why is it so?

Stupid Windowsisms (it worked everywhere else)

[Justin]

Originally Posted by tonalstates

Just tried it and wow, snappy! and I'm getting more results than before, thank you guys for this.

hmm more results? not sure it should change the output...

[tonalstates]

Originally Posted by Justin

hmm more results? not sure it should change the output...

Before, it seemed that reaper didn't search through all my subfolders, and now it seems it does (?) probably user error or I'm confused, speed was there though hahah I'll do some tests tomorrow on that thought

[musicbynumbers]

Originally Posted by schwa

btw this could/should be a significant improvement (factor of 10 speedup in our tests with a database containing ~1 million files with metadata).

+ Media Explorer: improve search performance with very large databases [t=270722]

Thanks! will try it next week. Appreciate these boosts for sure!

[BPBaker]

Originally Posted by tonalstates

v6.70+dev1117 - November 17 2022[*]+ ReaVerb: support deconvolution of more than 2 channels [t=272747] [p=2600567]

Amazing, thanks! I'm just in the first few minutes of testing it but I think I've successfully made a true (first order) ambisonic IR with Reaverb. Excited to play more with this (and experiment with higher orders).

[vitalker]

Originally Posted by Justin

Stupid Windowsisms (it worked everywhere else)

Bill Gates, what have you done?

[EvilDragon]

Maybe makes sense to call it : pow rather than : sqr then, if it can be overriden with other powers?

[BartR]

Originally Posted by tonalstates

[*]+ Web interface: fix fancier.html record armed track detection with a single track [t=272428]

Thank you it works greatly

[thommazk]

Thanks for the media explorer improvements!
While you're still making changes on this subject, could you take a look at this bug?

https://forum.cockos.com/showthread.php?t=270600

When media explorer is docked and you close/open the window, it resets the current search and file/database selected.

[mschnell]

Originally Posted by tonalstates

• + JSFX: support :sqr fader scaling, default is square, can override power with :sqr=3 e.g.

Maybe it's just me, but is square or ^3 really beneficial ? I would think exponential (i.e. dB) would be more useful. (Maybe with special handling of huge negative values to accommodate -inf dB)

-Michael

[Meo-Ada Mespotine]

Since JSFX support now 256-sliders, could this be upped for Video-processor too? Even if only a handful are visible, more sliders would be helpful for more advanced video-fx.

I would also like to see a slider-view, like JSFX have, which would be easier to navigate, than the current knobs, but that qualifies probably as its own FR...

[Justin]

Originally Posted by mschnell

[/LIST]Maybe it's just me, but is square or ^3 really beneficial ? I would think exponential (i.e. dB) would be more useful. (Maybe with special handling of huge negative values to accommodate -inf dB)

-Michael

sqr=4 is pretty nice imo

[vitalker]

Originally Posted by Meo-Ada Mespotine

I would also like to see a slider-view, like JSFX have, which would be easier to navigate, than the current knobs, but that qualifies probably as its own FR...

At least allow us to zoom knobs/their values with wheel, because it is not only hard to see them, but also hard to adjust on OS with >100% zoom.

[plush2]

[*]+ ReaVerb: support deconvolution of more than 2 channels [t=272747] [p=2600567]

After a quick test this one seems fixed here.

[BartR]

Originally Posted by Meo-Ada Mespotine

Since JSFX support now 256-sliders, could this be upped for Video-processor too? Even if only a handful are visible, more sliders would be helpful for more advanced video-fx.

I would also like to see a slider-view, like JSFX have, which would be easier to navigate, than the current knobs, but that qualifies probably as its own FR...

IMHO that processor should be improved in all the ways. But I understand this is not a priority. One day I think it will be amazing to have it more user-friendly (for Video I do use something Else, but in certain cases, certain users can get benefit from it, for simple montages, where the machine's power is a constraint)

[Phazma]

Originally Posted by tonalstates

[*]+ JSFX: support :sqr fader scaling, default is square, can override power with :sqr=3 e.g.

Excuse my ignorance but to what range and power would one set a fader to mimic a volume fader?

Originally Posted by tonalstates

[*]+ Media explorer: support zoom/scroll via mousewheel or horizontal mousewheel

Nice! How about allowing customization of the scroll/zoom behavior by adding mousewheel actions to the media explorer action list?

Also regarding mousewheel zoom/scroll, any chance this FR might be looked at in not too distant future? https://forum.cockos.com/showthread.php?t=251592

[Justin]

Originally Posted by Phazma

Excuse my ignorance but to what range and power would one set a fader to mimic a volume fader?

Eh, well, the REAPER faders use more complex model.. but :sqr=2.3 works pretty well for volume, e.g.

slider1:0<-120,12,0.1:sqr=2.3>volume (dB)

[mschnell]

Originally Posted by Justin

sqr=4 is pretty nice imo
...
:sqr=2.3 works pretty well for volume

Being a purist in math, I put a lot of effort in the Midi Volume JSFX on that behalf

-Michael

[TonE]

Originally Posted by Justin

Eh, well, the REAPER faders use more complex model.. but :sqr=2.3 works pretty well for volume, e.g.

slider1:0<-120,12,0.1:sqr=2.3>volume (dB)

For me any jsfx volume faders always need to use at max 0 dB, never more, not to destroy the original fine adjustments. I can set the exact range with the normal volume fader of the track, staying there statically. For any dynamic changes using instead jsfx vol_pan, for example, but only for cutting, never boosting. This is how I like it. This makes sure I will never destroy any fine made adjustment before. Why am I doing it, if I will destroy it any way, via any jsfx at some point.

The fine adjustments of volume I am changing using actions, in steps of 1dB or 0.5dB, or using a wheel controller in finer steps. The good side of the action based track volume change is there is no limit how far I can boost. Boost as much as you need, no problem.

So main idea is separating boosting (track vol) and cutting (vol_pan jsfx, for example).

[ErBird]

Originally Posted by Justin

Eh, well, the REAPER faders use more complex model.. but :sqr=2.3 works pretty well for volume, e.g.

slider1:0<-120,12,0.1:sqr=2.3>volume (dB)

Can the scaling please be defined in terms of the midpoint value rather than sqr= which is a bit ambiguous?

Another issue of sqr= is the shape of the scaling is skewed on one side of "linear".

Here, for m < 0.5, the shape is nice and what you'd expect from a scaling function.
For m > 0.5, not so much.

Polynomial



To solve this you can go piecewise:


Piecewise Polynomial

[ErBird]

Even better is to use an exponential. This will give perfect scaling on something like a frequency slider.

The only issue being that function falls apart when m = 0.5. But that's easy to account for.


Exponential

[ErBird]

Scaling based on min, max and midpoint curvature
https://www.desmos.com/calculator/pwne7tzboa



Scaling based on min, mid, max values
(More dangerous)
https://www.desmos.com/calculator/amo3v4tzkv

[Phazma]

Originally Posted by Justin

Eh, well, the REAPER faders use more complex model.. but :sqr=2.3 works pretty well for volume, e.g.

slider1:0<-120,12,0.1:sqr=2.3>volume (dB)

Thanks. I tried this and it works pretty well half way through an automation but towards the lower half the volume goes down pretty rapidly if using a value that goes far into the negative range, like -120.

[Tale]

Originally Posted by ErBird

Even better is to use an exponential. This will give perfect scaling on something like a frequency slider.

The only issue being that function falls apart when m = 0.5. But that's easy to account for.

I think that's the same formula I suggested here. And I agree it would've been better, but of course that's just IMHO, YMMV.

[Justin]

Originally Posted by Tale

I think that's the same formula I suggested here. And I agree it would've been better, but of course that's just IMHO, YMMV.

I suppose it's easy enough to add all of the modes (just a few lines). This one has the downside that it requires a normalized range so changing the range on an exported envelope will cause the curves to change.. but useful anyway (changing the ranges is probably uncommon).

[ErBird]

Originally Posted by Justin

This one has the downside that it requires a normalized range so changing the range on an exported envelope will cause the curves to change.. but useful anyway (changing the ranges is probably uncommon).

I'm really curious what you mean by this. You mean the input range of 0<-->1. That's not strictly necessary.
I mean, divide the input by 2 and now your input range is 0<-->2.

(exp(ax)-1)/(exp(a)-1) is just horizontal and vertical scaling an exponential.
It would just be nice if it can be done using a midpoint value like in the example I posted.
In my humble opinion that's is the gold-standard way to scale a parameter (aka match at 3 points).
Log is less friendly in the way it passes through the 3 points. Polynomial scaling is reasonably good.


I assumed Reaper's sliders operate on a normalized input range and just translate that to a scaled value.
Perhaps my misunderstanding is related to what I consider unexpected behavior when trying to link a linear and sqr slider.

If I parameter-link a sqr=2.3 slider with range -100<-->0
to follow a linear slider of range 0<-->100
the two sliders remain locked vertically locked for all slider1 values. This is what I expected for any slider2 range.




But if I change slider2's bounds to -99<->1 the linking of the two sliders is completely thrown off.

slider1=50 is mapped to slider2=0

so half of slider1's range controls -99<-->0 of slider2
while the other half of slider1's range controls only 0<-->1

This seems odd to me.

[ErBird]

Originally Posted by Tale

I think that's the same formula I suggested here. And I agree it would've been better, but of course that's just IMHO, YMMV.

It is effectively the same, just more directly useful.
Setting m is a lot more intuitive than trying to figure out an a value.

[ErBird]

More:

I was very surprised to see how scaling is being handled when the slider bounds cross through zero.

Red is expected, Black is what Reaper appears to be doing:
sqr=2.3 (no reason I'm sticking with that value)




To get to these formulas I measured the value of slider2 when vertically aligned with slider1 at regular spacing (0,10,20,...)
for slider2 ranges of -100<-->0,-90<-->10,...


Is this correct? The polynomial is split into two when it crosses zero?

[Tale]

Originally Posted by ErBird

It is effectively the same, just more directly useful.
Setting m is a lot more intuitive than trying to figure out an a value.

I guess it depends on what you're used to, but it's actually not uncommon to refer to a exponential curve as e.g. "exp 6.5" (where a = 6.5).

[AZpercussion]

ErBird, what math program do you use on the gif?
Thanks and sorry for offtop!

[nofish]

^ https://www.desmos.com/calculator

[ErBird]

Originally Posted by Justin

I suppose it's easy enough to add all of the modes (just a few lines). This one has the downside that it requires a normalized range so changing the range on an exported envelope will cause the curves to change.. but useful anyway (changing the ranges is probably uncommon).

Is this a more useable form perhaps?

https://www.desmos.com/calculator/s7shuo27lx



Maybe it has to be limited further so that x1=y1 and x3=y3.
As before, m can be defined as (y2-y1)/(y3-y1)

I've always dreamt of 3-point exponential sliders like:
slider1:0<min,mid,max,inc>

For example:
slider1:0<0,0.5,100,0.01>Attack (ms)
slider3:0<20,632.455,20000,0.1>Frequency (Hz)

[Justin]

Originally Posted by ErBird

Is this a more useable form perhaps?

https://www.desmos.com/calculator/s7shuo27lx



Maybe it has to be limited further so that x1=y1 and x3=y3.
As before, m can be defined as (y2-y1)/(y3-y1)

I've always dreamt of 3-point exponential sliders like:
slider1:0<min,mid,max,inc>

For example:
slider1:0<0,0.5,100,0.01>Attack (ms)
slider3:0<20,632.455,20000,0.1>Frequency (Hz)

hmm yeah that's nice. and the math can be simplified a bit since e^(2*ln(x)) = x^2, reduces down to ((m-1)/m)^2... not that it's that complex to begin with. we'll do that, :exp=2000 or whatever for the midpoint.

I'm really curious what you mean by this. You mean the input range of 0<-->1. That's not strictly necessary.
I mean, divide the input by 2 and now your input range is 0<-->2.

The advantage of the sqr/log modes is that the values encoded are the square/log of the actual values, so if you had an envelope with these values, and later modded the JSFX to increase the range of the parameter, those encoded values would still be correct. A pointless corner case, probably.