/ by Jan
Hi there, this is Jan Morgenstern writing from a currently particularly inhospitable part of northern Germany. I’m responsible for all things going “bonk” in Elephants Dream, and as some of you expressed their interest in occasional digressions into sound design and compositional topics in this blog, I’ve written a little blurb on SFX design for a scene near the beginning of the movie.
In this scene, the action takes place between giant telephone switchboards from the pre-digital era (such as these). Between the switchboards, long wires are frantically buzzing through the air, constantly connecting and disconnecting from the boards. The setting itself is part of an infinite machine, so there are no walls, just endless rows of weird machinery in the distance.
The Paranoia Simulator
As the wire action between the switchboards will be one of the defining elements of this scene, it was clear that it’d need a distinctive sound that would later take the most prominent part in the soundscape (apart from the dialogue, which has top priority at all times). Bassam had the idea of having the wires “leak” short snippets of phone conversations that surround the characters at all times, constantly moving around them or even chasing them, and thus creating a diffuse, surreal, and threatening atmosphere that envelopes our heroes.
In order to find appropriate source material for this effect, I rummaged through the web in search of some public domain voice recordings. Half an
hour later, I had recordings of some presidential speeches, old proceedings from US courts, and talks from the last Blender conference on my HD, and glued them together in order to create a single, large audio file of about 1 hour of playing time. I might end up using different source material in the finished movie, as we’ve been considering using recordings of our actors (maybe reading dutch christmas poems? ;)
As I played around with the voices, it quickly turned out to be rather tedious to create the desired illusion of movement manually, since the levels, panorama positions, and pitches of the voices had to be manipulated in unison in order to create a halfway convincing effect. Besides, as I only had an early animatic version of the scene as a reference, changes in the visuals would probably have made it necessary to throw large parts away and re-create them from scratch later. So I decided to delegate the dirty work to an automated signal chain I created in a software called Reaktor, which is a modular DSP environment. Reaktor can be thought of as a kind of virtual electronics breadboard in that it allows its user to build signal processing structures for almost every conceivable kind of purpose by interconnecting modules - ranging from simple arithmetic operators through various types of oscillators, modulators and filters to I/O modules.
At the core of the signal chain (the affectionally-named Paranoia Simulator) is a sample player that gets initialized to start playing my source file at a random point in time. In order to introduce some additional chaos, the initial playback pitch and level are slightly varied in a random fashion. The audio is then passed through a bandpass filter, once again with a randomized center frequency, to simulate a lo-fi transmission.
This signal is then fed to a substructure that simulates a sound source’s movement along a horizontal plane, along with the resulting level attenuation, time shifts, and doppler effects. In the original design, it also placed the source within the stereo panorama according to different microphone setups, but as I needed 5-channel surround output this time, I left this part out and fed the resulting (mono) signal to a surround panorama module instead, which in turn feeds the simulator’s audio outputs.
The movement itself is created with a simple Lissajous generator, which
generates 2-dimensional motion patterns based on sine functions. Once
again, the parameters are constantly varied randomly in order to have the voices move around the listener in a most unpredictable way. This generator can easily be replaced with a user control in order to provide manual control of the movement. This way, I can quickly add some layers of deterministic movement against a randomized backdrop if some of the wires are doing anything relevant in the foreground.
Enough yakking, let’s hear this doohickey! :)
This is the output from a single processing strand, with no further
surrounded_by_bassam.mp3 (Stereo, 812kB)
For the final effect, I increased the voice count to a larger number
(10 in this case) and added a bit of reverb:
voices.mp3 (Stereo, 984kB)
voices.ac3 (5.1 Surround, 1.6MB)
Jack in (and out of) the box
There are still some things happening in the visuals that don’t have an acoustic counterpart yet. For example, plugging jacks into and out of a switchboard makes a distinct clicking sound. I recorded plugging an audio jack into the rear of an old outboard unit, loaded the resulting sounds into a sampler, assigned them to different keys on the keyboard, and triggered them randomly:
jacks.mp3 (Stereo, 420kB)
For the surround mix, I simply added a second stereo layer of clicking sounds at a lower level (to avoid drawing the viewer’s attention away from the screen) to the rear speakers.
The soundscape is already getting a little crowded at this point, so I’ve tried to keep additional sonic elements more or less in the background; also, since there’s already a lot of movement, I preferred the remaining sounds to stay at static positions in order to provide some anchoring. As the characters roam freely around a large industrial environment, there’ll be omnipresent sounds of machine parts everywhere. Since the purpose of the machine is not clear at this point, the environment sounds should be somewhat diffuse in character and should not point to a single, obvious origin, so I’ve mostly resorted to a rather eclectic mix of mechanical, hydraulic, electric, purely synthetic, and organic sounds.
Putting it all together
Now I had separate stems of the 3 elements I described on my HD, so I could go ahead and mix them in order to get the final SFX track for this scene. To enhance the feeling of vast space, I’ve added some reverb to all elements - as the viewer’s focus rests on the wires and jacks in the foreground, those got less reverb and more direct sound, and vice versa for the background machine parts. In a last step, I’ve slightly attenuated the frequencies between 500Hz to 2.5kHz in order to make sure the dialogue won’t get disturbed by other elements (this frequency range is most critical to speech comprehensibility). This is a somewhat preliminary measure, as I suspect there’ll be plenty of things that’ll need some additional tweaking in the final mix.
This is the semi-finished SFX track for the scene:
soundscape_final.mp3 (Stereo, 1.4MB)
soundscape_final.ac3 (5.1 Surround, 2.1MB)
That’s it! Of course, I appreciate any feedback (like “I made it to the third paragraph before falling asleep and/or into a coma!”). I’d also love to hear from people with surround rigs hooked up to their computers out there (if any) whether the AC3 files work for them.