Creating Three-Dimensional Computer Animations Using Spectral Data and OpenGLSkip other details (including permanent urls, DOI, citation information)
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License. Please contact email@example.com to use this work in a way not covered by the license. :
For more information, read Michigan Publishing's access and usage policy.
Page 00000097 Creating Three-Dimensional Computer Animations Using Spectral Data and OpenGL Tim Kreger Australian Centre for the Arts & Technology Institute of the Arts Australian National University Abstract The author has been exploring the mapping of spectral data to the visual domain for artistic purposes. The technique involves analysing sound and music using Fourier techniques and using the data to control an OpenGL animation environment The result is an "automatically" generated animation which is synchronised with the original sound file. This gives the composer of electroacoustic music a relatively quick and simple way to produce "video accompaniments" to their work should they wish to. The process raises many interesting problems and issues about the perceptual relationship between music and image. These issues will be discussed and examples of the work-to-date will be presented. 1 Tools 2 Process 1.1 OpenGL OpenGL is fast becoming the standard API for threedimensional graphics development as it provides a streamlined, hardware-independant interface to graphics hardware on most commercial platforms. The programmer requires no knowledge or understanding of the host machine's graphics hardware as the OpenGL libraries on- that machine provide the neccesary optimisation and hardware calls to ensure the graphics are rendered as quickly and as efficiently as possible. The Macintosh version of this software actually uses MacMesa, a freeware 3-D graphics development API which is syntactically identical to the OpenGL libraries and is available on most popular platforms. 1.2 FFTW FFTW was used for analysis as it provides a quick, easy to use enviroment for doing Fourier analysis. FFTW doesn't require the number points in the analysis to be a power of two which can be useful in the context of visualising sound data. The libraries are available on most popular platforms and are licensed under a GNU licensing agreement The technique employed involves analysing a sound file using the FFTW library and using the magnitude values of each bin to control primitive objects in the OpenGL animation. One analysis is done per frame of animation. The number of analysis points is arbitrary and determines how many primitive 3-D objects will appear in the animation. Some isues arise when the frame size of the analysis is significantly smaller than the sample rate divided by the animations number of frames per second as important sonic events may be missed by the analysis and hence not appear in the animation. In practice this seems not to be too problematic as the ear-eye-brain percept is not acute enough to notice these inconsistencies. On each analysis and visualisation the visual frame is stored into a SGI-type RGB image file which is later compiled into a Quicktime, AVI or whatever movie format is required. It should be noted that the sound is resynched to the image in post-production. This approach was taken so as to increase the flexabilty of the process in the context of producing animations. 2.1 Mapping Although the process is relative simple, much thought has gone into how the 3-D objects are placed in the space of the visual domain. Initial emperiments use the magnitude parameters of the analysis to control the ICMC Proceedings 1999 -97 -
Page 00000098 radius of a sphere, one sphere per analysis band. The spheres are laid out along a line sequential in accordance with their position in the analysis. This approach works well in terms of seeing spectral activity but is visually quite boring. Also if one uses a large analysis frame then it is difficult to see all of the spheres dues to the limited real-estate of the graphics screen. The spheres will either be too small to see any detail or large enough to see detail but not enough of the spheres can be seen so the entire spectrum is not visible. This is the area which is recieving the most research at the moment and will be demonstrated during the session. 3 Software Parameters 3.1 Input Settings The input settings are used to set the input sound file and the number of bands in the analysis. The sound file must be a raw 32 bit floating-point file. 3.2 Amplitude Settings The amplitude parameters are perhaps the most important for achieving a meangiful animation. These parameters determine how the analysis information will be used to control-the primitives. The Amplitude Scalar is multiplied by the magnitude component of each of the bands in the analysis so that the primitives will appear at the optimum size for viewing, this is very different for each sound file and takes some time to get right for each sound. The Amplitude Threshold is used to reduce visual noise and rendering time by making sure that only spectral components with amplitudes above this value are drawn. The Amplitude Clamp is used to limit the upper size of the primitives. Without this paramater some primitives will overly dominate the image, particularily if the Amplitude Scalar is high. 3.3 Texture Settings The texture parameters determine whether the primitives are to be texture maps and if so how the texture is to be applied. The texture has to be an RGB image file. One can use an animated sequence of RGB's as the texture which make it posible to use dynamically changing textures. There.at three texture modes in the OpenGL enviroment: Object Linear is when the texture is mapped onto the object so that it moves with the object surface. Eye Linear is when the texture remains in a fixed position with repect to the viewer. Sphere Map is when the a spherical model of the texture is mapped on the entire enviroment. This is useful for simulating reflective surfaces. 3.4 Camera Settings The camera settings are used to position the camera within the space which is useful for positioning the camera in the optimum space for each animation. It is hoped that these parameters will be scriptable at a later date so that the camera can be animated. 3.5 Light Settings The light settings are used to position the light with the space. This parameter needs to be used in conjunction with the camera postion so that the model will appear in the rendered image. 3.6 Render Settings The render settings are where the frame rate, render range and name of the output files are set. The frame rate is useful for running quick tests at lower frame rates to ensure all other parameters are set correctly. The render range facility makes it possible to produce animation in smaller chunks enabling one to make animations from larger sound files when disk space is an issue. The software outputs a sequence of RGB images with names in the format filenameXXXX where XXXX is the number of that file with respect to the sequence. 4 Conclusion The process is so far quite a simple one and the author hopes to develop this framework into a larger more felxible enviroment for creating animations. It is envisaged that it will provide a tool for composers to explore the posibilty of producing visual accompaniments to their music or sound work. - 98 - ICMC Proceedings 1999
Page 00000099 5 Acknowledgements References The author would like to thank Larry Polansky and the Bregman Electronic Music Studios at Dartmouth College, USA for providing the opppurtunity to cary out this work. The majority of this work was done during a a residency at the above mentioned facility. Neider. J, Davis. T, Woo. M, OpenGL Programming Guide:the official guide to learning OpenGL, Addison Wesley, New York 1993. Roads, C. Computer Music Tutorial. MIT Press, Cambridge, Massachusetts 1996. ICMC Proceedings 1999 - 99 -