Cloudbounce Blind Test Results

 

JPEG image-32A07E328359-1

My mixing/mastering DAW of choice, Tracktion, has just added CloudBounce as a feature.  This is one of those online mastering services like Landr, centered around a machine-learning algorithm.  The results of the blind test are below.

 

I tested Landr also when it launched and the results indicated it was expecting a more conventional kind of track – not the variable dynamics and lack of consistent transients common to electroacoustic improvisation (or “noise” if you’d prefer!) While i’m happy to explore procedural generation as a creative tool for visual art and sound design i’m not entirely convinced it has a place in finalisation of tracks as these choices seem to me to be highly personal.

JPEG image-FC54F14E4F2F-1

 

Having said all the above – while I’ve studied some of the approaches to mastering – I have no specific qualifications.  So I thought i’d do a blind test.  The playlist below features a track mixed by Joe from an session we did a month ago.  I’ve uploaded his raw stereo file to Cloudbounce to see what it could make of it.  I’ve also done my own mastered version in the box using only the software I have regular access to.  The results can be streamed or downloaded.

My questions to listeners:

  1. Which is the Cloudbounce master and which is my version?
  2. Which do you prefer?

RESULTS (including poll data and PMs)

Note very few people who PM’d me chose to guess which was the CloudBounce master however of those that did, and including the poll results, 60% of the vote guessed the second track was mastered through CloudBounce.

The reality is that the CloudBounce master was track 1.

The response was pretty unanimous that track 1 was the preferred track with greater definition.  It was also considered to be “louder”.  One user responded that they thought the second track was “dull” whereas another, while suggesting track 1 was a better mix, liked the softer focus of track 2.

Personally I was pleasantly surprised at the quality of the CloudBounce master. Of particular interest to me was the increased sonic space it brought out, situating the spiky transients more effectively amongst the deep / distant reverberant sounds.

One caveat is that my mastering attempt was not thorough.  I was working with a pre-mixed stereo track and merely applied Airwindows Console4(Buss and Console) with Melda Spectral Dynamics and TurboComp flattening the overall sound somewhat which perhaps accounts for the dull / unfocused result.  Clearly I have some learning to do.

I’m certainly keen to use CloudBounce services again.  If you are reading this and would like to test it out please click this referral link and let me know how you go.

 

 

Not all people like what I like and that’s just fine!

Here is a musical demo combining Push, Madrona Lab’s Kaivo and Zynaptiq’s new Adaptiverb which sounds like it could be an essential tool for us Ambient muso’s who use reverb as an instrument.

I haven’t been super productive over the last fortnight as I’ve been spending most of my spare time with No Man’s Sky, the latest implementation of a thing I very much like, exploring procedurally generated worlds.

NMS_Mania
My ugly mug in a NMS landscape via Pikazo

Noctis is probably the earliest obsession in this regard though i’ve always played flight simulators and open world RPGs for much the same reason.  While there is an overarching story to the universe it is mostly flavour to set atmosphere.

NoctisIV
Noctis IV

I showed NMS to Joe and he agreed it was “Noctis on ‘Roids”.

nms_example
No Man’s Sky – can’t capture screenshots without Steam – very tech-no-logic-al.

What has been slightly depressing is watching the online tantrums among those who were clearly expecting something more like Star Wars.   Game elements are minimal and repetitive but comparable to similar titles with a similar focus on exploration and procedural generation.  While it is certainly buggy and unoptimised, something that is more glaring is the apparent disparity between what was demonstrated during development, feeding the hype, and reality.  This amusing video summarises it quite well.

It’s quite clear to me the first one does not utilise substantial procedural generation as the creatures look realistic… and dull.

I recall the same negative response to Spore – and this excellent article by Soren Johnson thoughtfully explores the chasm between ground-breaking theories and hard truths about game design.  I certainly think that NMS is an interesting case.  While it didn’t suffer the same amount of interference with a much smaller development team it certainly seems to be copping a similar backlash.

Library of Babel
The Library of Babel

One thing with Procedural Generation is that you can end up with the “Library of Babel” effect – illustrated above by a generated page from one book, on one shelf, on one wall of the section in the virtual library generated by my birthdate in hexadecimal format.  A real page-turner!

If the player can bring some of their own imagination to NMS then I feel it is quite an engaging procedural sandbox but placing this in a commercial game is clearly problematic.  Noctis IV was a free game as is the more serious / less gamey Space Engine.   Artmatic Voyager (from the makers of Metasynth) is a payware title that generates Procedural Worlds but not real time environments; rather it renders attractive backdrops for you to populate with your own sci-fi assets.

artmaticExample
Artmatic Voyager

Perhaps the most promising procedural worlds generator on the horizon is Ultima Ratio Regum – but that appeals more to the rogue-like obsessed who also worship at the altar of Dwarf Fortress (nothing wrong with that!)

Anyway – that’s my two cents.  To return to the more music focused purpose of this blog – Ohmwrld is something I worked on with Joe that is inspired by virtual worlds and features cover art generated by Bryce, one of the earlier strands of Artmatic Voyager.

And here is something that sounds procedurally generated (but is actually improvised) that I made recently with Paul Forbes-Mitchell.

Final thoughts – the death of Makrotulpa is exaggerated – something new is coming.