I went to a meeting on Friday to discuss the development of a standard for describing spatial audio, called SpatDIF. There is now an official website and mailing list, so go check it out at www.spatdif.org.

The idea is quite nice, because it allows one to separate the contents of the 3D scene from the audio renderer. All that is required is to format messages for sound source positions and directivity in this OSC-style format:

/spatdif/source/1/position 1.23 7.98 5.4
/spatdif/source/*/directivity 1 0.7 0 0.7

Then, any SpatDIF-compatible renderer will be able to spatialize the sounds on the particular audio display that it is written for (eg, binaural, 5.1 surround, VBAP, Ambisonics, etc).

I’ve finally decided to consider standardized formats for 3D content (re: virtual worlds), or for that matter, the lack of… Since there doesn’t seem to be any consensus, except maybe for the Web3D Consortium, which has embraced X3D (the successor to VRML) as the standard for interchange of 3D scenes, especially for the future of ‘3D’ web browsing.

As a side-note, it seems like Collada is fast becoming the standard interchange format for 3D models. This is not the same thing that I’m interested in. I have little interest in describing 3D contours in an interchangeable format; rather, I’d like to have a way to describe vast virtual worlds so that they can be rendered on many clients, particularly with support for Linux and OSX.

So, I came across two promising libraries: OpenVRML and X3DToolkit.

Here is how I built X3DToolkit one OSX 10.5:

1
2
3
4
5

$ cvs -d:pserver:anonymous@x3dtoolkit.cvs.sourceforge.net:/cvsroot/x3dtoolkit login
$ cvs -z3 -d:pserver:anonymous@x3dtoolkit.cvs.sourceforge.net:/cvsroot/x3dtoolkit co -P x3dtoolkit
$ cd x3dtoolkit/src
$ qmake
$ make