2CAudio Kaleidoscope | It's A Trip | Latest Update 1.1

[Andrew Souter]

We should be posting a new Win version for existing customers to try later today!

Working on getting the OSX version ready now too. Next week for OSX I assume unless something unforeseen happens...

We will consider these public betas (for existing customers). They have been tested extensively by us on our machines, and are now ready to be tried in the real world. Our standard testers will receive them at the same time. If nobody finds any major issues these will likely become the final versions for this update cycle. We'll see. We'll send an official announcement after a few weeks in the wild for anyone who would like to try it.

Performance should be drastically better for most presets and hardware systems, particularly for Spring resonators. At large buffer sizes you might see more than 3x as fast/efficient/light regarding CPU usage. Sometimes even more.

Rendering the second factory preset "Ancient Transmission 01" for 5 min of audio for example used to take 48 sec on my 12-core mac pro. It can now be done is 7-8 seconds in some cases!!! Which should also translate to meaning you could run ~5-6 times as many of these real-time as previously for this particular preset... This is an extreme example, but it is not entirely uncommon.

In sum it is safe to say, things are A LOT faster in most cases!!

This update focuses on three things: performance, performance, and performance. We did sneak in a couple, but powerful, new features though...

more later this evening.

[lnikj]

How very dare you give it to Windows users first!

[SJ_Digriz]

can't wait!!! Thanks for the news.

[Andrew Souter]

How very dare you give it to Windows users first!

We develop in Visual Studio on Win primarily. OSX port is usually the last step sorry to say, and I type this on a Mac, and do my on music production on OSX, so I empathize, believe me.

But since win is ready, we thought we'd get it out there to get extra feedback while finishing the OSX build.

[shroom81]

Nice one Can't wait to try it out

[Aiynzahev]

Thank you Andrew

I can't wait to try it

[Andrew Souter]

OK Windows customers, whoever would like to try it, give it a go:

http://www.designersound.com/product_in ... cts_id=116

Follow the directions there using the coupon code.

As a precaution please copy your user presets somewhere before running the installer. This should be unnecessary, but better safe than sorry. User presets should be in:

C:\Users\*You*\Documents\Kaleidoscope\Presets\Full\User

(and please verify for me that the installer does what it is supposed to do and retains your user presets, and thus this step is unnecessary.)

This is a public beta version for customers who would like to try it and give feedback. If you are not comfortable with such things, please wait until the official announcement.

[Andrew Souter]

Kaleidoscope 1.1.0 Update Release Notes (For Win Customer Beta):

Changes:

1) MAJOR PERFORMANCE EHANCEMENTS

• Efficiency should be at least 200-300% better and many cases significantly more than that when using the latest generation CPUs.
• Efficiency is most improved for Spring resonators. String resonators are improved by 50-75% or so, before considering the other significant performance enhancements below.

2) RESONATOR THRESHOLD PREFERENCE

• Max Enabled Lines preference has been replaced with "Resonator Threshold". This is a much more intelligent scheme that disables resonators when their gain is below the specified threshold. This allows resonators which have very low volume and are inaudible or close to it, to be disabled, thus saving additional CPU resoruces.
• This setting is independant for Real-time and Offline, and a fairly high Threshold option of -48dB is provided to use on lower power CPUs. This can save a huge amount of CPU resources when dealing with presets that have very sparse images, and/or extreme Soft values (i.e. low amplitude of high frequencies.) Since the disabled resonators are extremely low in gain their contribution to the total sound is reasonably minor, and more aggressive (higher) threshold settings can be used for real-time to save additional CPU resources, while the rendered result can include the resonators that are disabled for realtime by using a lower threshold value.
• The 200-300% figure above is NOT including the use of this new preference, and is a fair test comparing equal active lines between KS 1.0.2 and KS 1.1.0. When using a threshold setting of -48 dB, many preset settings can see even more drastic speedups. Factory preset "Ancient Transmission 01" for example is almost 700% more efficient in KS 1.1.0 compared to 1.0.2!! SEVEN HUNDRED PERCENT!!
• Increases to efficiency via using Resonator Threshold apply equally for all resonator models, therefore many string presets may see speedups of a few hundred precent as well.

3) NEW BUFFER OPTIONS

• Kaleidoscope's buffering system has been completely redesigned.
• There are independant settings now for Real-Time and Offline. Offline Settings should generally be set to the maximum value which is currently an extremely large value of 65536 samples. Large buffers increase processing efficiency, and since we do not care about latency during offline renders, we can use an extreme setting to increase performance. This is particularly helpful for large multi-core systems.
• Realtime Buffer options now include "Auto". When using Auto KS does not use any additional internal buffer, and uses only the host/hardware buffer. In this mode KS adds ZERO ADDITIONAL LATENCY. The "Auto" mode should generally also be the most efficient for real-time use. (* Host buffers should still be set to a large value for maximum performance.) "Auto" should be the default choice for Real-time use. We may remove the other options and retain only Auto, depending on user reports/requests.
• Kaleidoscope works with any size host/hardware buffer including non-power-of-two sizes as are common with some Firewire devices.

4) FULL DOUBLE PRECISION END TO END

• Kaleidoscope is now a completely "double precision" process. The entire algorithm from input to oversampling, to all processing, to downsampling, to output is done entirely in 64-bit precision math. This is a scientific level of precision and results in a noise floor of something ridiculously low around -300dB or so. Kaleidoscope processing is effectively mathematically perfect.
• Not only is Kaleidoscope much faster, it is also more precise, which is usually considered an impossible task in engineering.

5) NEW FEATURE ENHANCEMENTS

• Feedback Relativity has a new bipolar range. Negative values increasingly reduce the decay time for high frequencies making them less resonant. This can help model various natural materials such as wood and biological materials where high frequencies do not generally resonate/ring very long.
• Mod Offset has a new bipolar range. Negative values will force some pixel values exactly to 0.0 (i.e. "black", i.e. "off"). This allows dense images with lots of non-black pixels to be make much more sparse. This produces very aesthetically interesting sonic results and allows "random picture" data to be a little more useful. It also has significantly performance implications when used together with the new Resonator Threshold preference, in that it will effectively disable more and more lines as their respective gains are reduced below the threshold.
• Mod Curve has several new interesting options including: Hyperbolic, Asymmetrical Bands, Warp, Iterations, Triangle, Chop, Dual, and Quantize.
• Combining the new Mod Offset range with the new Mod Curves produces and exponentially greater range of image transformation possibilities.

[Sampleconstruct]

I assume all presets we made are fully compatible with the new version so that we don't have to re-edit them in any way?!?!

[Sampleconstruct]

And still no alternate looping in version 1.1.0?

[Andrew Souter]

I assume all presets we made are fully compatible with the new version so that we don't have to re-edit them in any way?!?!

Yes, of course.

And I am sorry I don't have it to give to you first! You'd be first on my list to give it to try out if we had the OSX version to give. Hopefully not longer than sometime this coming week.

Since we more or less re-wrote the "guts" of KS completely, we want to extend the beta version to any customer who would like to try it... to be sure stuff like the buffering changes agrees with all the weird and wild hosts and hardware out there...

[Sampleconstruct]

I assume all presets we made are fully compatible with the new version so that we don't have to re-edit them in any way?!?!

Yes, of course.

And I am sorry I don't have it to give to you first! You'd be first on my list to give it to try out if we had the OSX version to give. Hopefully not longer than sometime this coming week.

Since we more or less re-wrote the "guts" of KS completely, we want to extend the beta version to any customer who would like to try it... to be sure stuff like the buffering changes agrees with all the weird and wild hosts and hardware out there...

Sure, no worries, I can easily wait for another week, I've been power-using KS heavily/every day for 15 months as it is now, so these performance issues were not affecting me so much. But I'm looking forward to the new features to hear, what kind of new sounds they will make possible.

[Andrew Souter]

And still no alternate looping in version 1.1.0?

we have something better in store for that...

this release is about the three Ps:

performance, performance, performance.

The other features are things we couldn't help ourselves from putting in there...

Features will come next, and won't take as long.

[Sampleconstruct]

And still no alternate looping in version 1.1.0?

we have something better in store for that...

Hyper-hyper-looping, warp-looping, 4D-looping, looping-looping, anti-looping, speed-of-light-looping?
Bring it on when you feel it's ready.

[Andrew Souter]

Sure, no worries, I can easily wait for another week, I've been power-using KS heavily/every day for 15 months as it is now, so these performance issues were not affecting me so much. But I'm looking forward to the new features to hear, what kind of new sounds they will make possible.

Well, imagine you can now basically probably do completely "live" ambient KS pieces with many more instances than previously... if you could get 15 or 16 of that test preset perviously, 30-40 real-world instances should likely be possible on your (and my) machine now... plus some B2s. Ambient heaven.