Sunday, May 30, 2010

Courvoisier released!



Update: Revision C released with fixed M/S matrix.

Download link
Download link #2

Courvoisier is a single-band universal dynamics processor with adaptive attack and release algorithms optimized for complex program material (mixes and submixes). Courvoisier uses free-form curves drawn over the histogram of incoming audio material (as opposed to traditional ratio and threshold parameters) to determine the transfer function of the processing.

Courvoisier is available for free.

Features:

- Histogram representation (sliding-window statistics)
of incoming program material and the resulting gain change
on which the transfer function can be drawn

- Free-form upward and downward compression and/or expansion

- Spline interpolated control curves

- All performance-critical code optimized in assembly
(utilizing SSE2 flow instructions)

- Two different adaptive attack algorithms (program dependent)

- Adaptive release algorithm (program dependent, scalable)

- Additional user defined adaptive release behavior

- Side chain filtering (low and high pass)

- MS operation
(with support for different transfer curves for Mid and Side channels)

- Stereo operation (adjustable stereo link)

- Dry mix option for parallel processing

- Peak, RMS and Hybrid level detection

- Very low harmonic distortion
(THD during normal operation is below -105 dB FS)

- Lookahead processing for minimizing distortion
with extremely fast attack settings

- Complex mode (oversampling with a twist)

17 comments:

Unknown said...

Hi,

Thanks for such amazing and interesting dynamics processor, it is very good to see someone cares about innovation :)

I've read the manual and tried it on different material: outstanding (though I'm pretty sure I'll get better results once I get deep more into it). I'm using Reaper and I noticed it has a Denormals issue when idle (it can take up to 20% of CPU on a Core i5). I just put a dither before and problem was solved (I thought it was worth mention it).

Thanks again for your time and generosity. I'll keep an eye on this, it'll be really useful on many situations where a full control of dynamics is needed.

P.S. - I'd like to contact you via mail next time, could you please provide me an address? My mail is mercadonegroestudios at gmail dot com

weldroid said...

Thanks for the kind comments and the feedback (especially on the denormal issue). I will look into that promptly.

Cheers,
Weldroid

spanimo said...

Nice Plugin, we have a review in german language on our music production site, with soundexamples.

Weldroid Courvoisier, erstes Review mit Klangbeispielen.

weldroid said...

Nice review spanimo, thanks for sharing!

Anonymous said...

I should admit that you make great job, weldroid! Thank you very much for it! With pleasure use your ambio.one and try to use Courvoisier. Sounds very smooth to me, I like it very much. Thanks!

What about auto-curve-adoptation which is mentioned in the manual? ;)

And what about low CPU support, is it possible for it?

It would be nice if you add Load, Save and Reset in the plugin for ex. like in Voxengo plugins by Aleksey Vaneev.

Also it would be great if you add remove dots function.

weldroid said...

Cheers Eduschka!

Well, adaptive release and attack mechanisms are so-to-say not visible directly on the GUI, but are working `under the hood` all the time.

Lowering CPU usage would be indeed possible, but it would require some major re-design of the adaptive algorithms (that`s what eats so much CPU), and might sound a bit different, so it may happen in the future, but not until a while.

Preset management for now is handled by the host, but thanks for the suggestion, I will consider it in a future release.

What do you mean by `remove dots function`?

Anonymous said...

Hello weldroid!

Thanks for you rapid answer to me. :)

You see, sometimes it is very needed that plugin could save preset itself.

According to the dots I mean dots on the grpaphic. When we change them we cann't remove them if they are not needed any more.

weldroid said...

Sure you can. Just drag them off the `screen` towards the top or bottom.

Anonymous said...

Thanks weldroid!

Maybe I did't read your manaul carefully. :*)

Anonymous said...

Weldroid, just try to use your Courvoisier with Refined Audiometrics Laboratory CLAS. It's absolutely FREE. You will receive simply unbelieveable results.

Cheers!

weldroid said...

Cool, I will try it. Glad that it works you wonders :-)

Unknown said...

Hello
In your Manuel sec. 4.2.2

On our next example the transfer curve has been changed from its default
flat shape. The curve has a dip around -19 dB FS, and therefore
Courvoisier will try to make levels around -19 dB FS quieter. The curve
also features a boost around -39 dB FS. Courvoisier will boost these –
relatively low – levels accordingly.
Figure 2: The transfer curve and the gain change histogram
The transfer curve has the vertical scale of +/- 12 dB, where 0 dB is in the
middle, and +12 dB is where the curve is touching the top of the screen.
The histogram of the resulting level change (representing the difference
between the input’s and output’s histogram) is now also visible (blue plot).
According to the statistics the processed output now contains more levels
around -15 dB and less around -26 dB.


Why is there more levels around -15 db and less around -26 db when you cut around -19 and boost around -39?

also the blue line does seem to move as much as the boost and cuts why is that? if i boost some 12db should not the blue line go up 12 db?

weldroid said...

"Why is there more levels around -15 db and less around -26 db when you cut around -19 and boost around -39?"

Because the combined effect on the distribution of the levels (blue plot) of attenuating levels at -19 dB and boosting at -39 dB will be that some of the lower levels will be higher (in crude terms those levels move from -39 dB to -15 dB). Also "cutting" at -19 dB will make those levels lower, hence they "move" down the chart.

Of course, one tends to interpret the charts of Courvoisier just like an EQ (you boost at one point along the X axis and so you expect to have a positive increase at that place), but because the underlying graphs present statistical information, it is more like one sets the dynamics processing target with the curves (based on the initial red histogram) and then observes the actual results (blue histogram). Takes a bit of getting used to.

weldroid said...

"also the blue line does seem to move as much as the boost and cuts why is that? if i boost some 12db should not the blue line go up 12 db?"

It should indeed, if Courvoisier was a waveshaper (that is: it would have an effective 0 ms attack and release time). You don't get 12 dB boost, because the adaptive attack and release rates put a limit on how fast the output gain can change, and so, the black curve defines a target that serves as a target for the gain (reduction/boost), but the actual gain will never reach that.

Courvoisier is optimized for transparency, and so it handles the output gain conservatively in order to minimze distortion (or unnatural gain flutcuations).

Unknown said...

Thank You Weldroid

I like Courvoisier, i just need to learn it better, I most likely use it on my next project.

I leave you with peace
EMCEE EXORCIST
WWW.MYSPACE.COM/EMCEEEXORCIST

weldroid said...

You are very welcome. Hope you will have much fun and great results!

Dax said...

Hi Weldroid,
I'd like to talk to you about developing this plugin a bit more. Could you please email me? dax2puzzlefactory>com>au
Thanks.

All the best,
Dax.