Despot - hard hitting compressor - testing

[dekadenz]

Oversampling will be optional of course and with selectable ratios from x1 to x16 (not sure though). First version with OS will have quite high latency (well known library without compromises for non-realtime use mostly).

[dekadenz]

New build, featuring linear phase oversampling (r8brain by Voxengo) is out.

Muted audio on ratio change is not a bug - most hosts (including Reaper) handles latency changes improperly, so consider this an invitation for restarting playback until I work around this and test in a few hosts at least. The same with release lesser than attack.

I recommend x4 oversampling minimum, x16 for ideally. Higher ratios are included for fun mostly - bring your CPU down on its knees.

https://stash.reaper.fm/v/24184/despot-1432884729.zip

[SaulT]

How does one go about building a zero-delay envelope follower? I started googling around and became interested in using Hilbert transforms for envelope detection, but that's not amenable to zero-delay/TPT.

I imagine that a prototype lowpass filter set at an almost DC cutoff would be amenable to zero-delay...?

[dekadenz]

Envelope follower is nonlinear one-pole lowpass filter. Nonlinear element instead of cutoff point. Design such a filter first (filters with nonlinear elements are described in Zavalishin's book) then solve the equation.

That's where the difficulty is. Either simple functions that will sound horrible or transcedental/high order polynomials that aren't easy to solve. And the solution won't be CPU-friendly.

UPD:
You really should switch to C/C++ if developing such a nonlinear ZDF filter. JSFX performance will be around 0.1x on recent Intel processor even for fairly simple ones.

There are some "cheating" pseudo-ZDF methods of course, but they aren't interest me. Hardcore only.

By the way, have written a script version with Hilbert RMS detection by FFT convolution some time ago for "internal" use and testing (long FIR filter is required, IIR pseudo-Hilbert or short FIR are useless and does not make any difference with peak detection or lowpass RMS). Probably, some of my distant future releases will have this as an alternative to peak detection.

[Joshua_Noncom]

Hi dekadenz,
Ok, this compressor is amazing. Just wanted to chime in and say Full support it is incredible! Love the ultra clean hardness without sounding cheap or brittle. Does things I try to do with L2, 1176, or Distressor and they never quite get it. This does! Also, using it in parallel is perfect. To my ear completely free of phase weirdness, or unwanted noise and distortion. Just a more dynamically limited version of what you put into it. I rendered a test compressing a drum machine part find it attached. 4bars dry, 4bars fade between 100% dry and 100% wet, and 4 bars 100% wet.
https://drive.google.com/file/d/0B9o...ew?usp=sharing

Fantastic piece of engineering! Thank you for going above and beyond.

[vanhaze]

Mac version ?! Please !!!

[citizenkeith]

Originally Posted by vanhaze

Mac version ?! Please !!!

Yes please!

[saastara]

This thread peaked my interrest but alas, no mac version. Spare for some salt for the wounds, is someone still using this one?

[SaulT]

dekadenz, I've sent you a PM.

[dekadenz]

Hey,
To everyone who expressed an interest to my unfinished magnum opus.

It seems that I've managed to get over my disappointment with IT area which caused my descent into the pit of depression a few years ago. It's royally screwed up indeed, still it's my one and only interest.

So, for the first and foremost, going to release the version 1.0 with a few minor bugfixes and improvements, including optimized oversampling (indeed, better than r8brain!).

Then, despite my dislike for crowdfunding model, I think it worth a shot, since I've amassed quite a lot of ideas during this time. The result will be freeware for everyone, no commercial versions. Functionality will be gradually improved after reaching an each funding goal, in other words, next goal = next version.

Whaddayasay?