Web3D RoundUp:†Looking Backwards and Forwards

Vol.34 No.2 May 2000

VRML Dream On-Line Performance

Stephen N. Matsuba
VR Studios

Figure 1: A screen capture from the original†VRML Dream†performance.

Figure 2: A screen capture from the motion capture trials presented at the Web3D RoundUP in Paderborn, Germany.

In the late summer of 1997, Stephen N. Matsuba and Bernie Roehl began an ambitious project: to create a new entertainment medium, without funding and together with a dedicated group of volunteers situated all around the world. The project sought to present a live performance of Shakespeareís A Midsummer Nightís Dream using real-time 3D animation multicast over the Internet. The sets, props and all the characters are optimized VRML models. A group of actors provided the voices and eight "puppeteers" controlled the charactersí movements. The voice and the motion data were digitized, compressed and sent out over the Internet in real time. Moreover, people were able to access this multicast with a 28.8K modem connection and a 150 MHz Pentium computer. Less than six months after its inception, VRML Dream made "Internet history" [1] as the first live, streaming, entertainment project with a running time of more than two or three minutes.

Two key problems presented themselves in the development of this project. One concerned the 3D models for VRML Dream. The VRML Dream projectís primary goal was to be a viable theatrical production. According to the artistic statement, a key reason for using VRML was "to enhance particular themes inherent within the play itself" [2]. In other words, the medium also had to play a role in the artistic goals of the production. At the same time, however, the models also had to conform to the kinds of restrictions that are inherent in developing material for real-time 3D animation and the Internet. The characters and sets had to have low-enough polygon counts to allow for smooth animation on the target computer platform. Since the goal was to make VRML Dream accessible to people having only a standard 28.8K modem connection, the model designers also had to consider the file size of each model.

Another area of concern involved the software needed to produce VRML Dream. Much of the streaming and motion software had to be created by the project members. The programs that generated the motion data for the charactersí avatars (as well as the network and client-side software) were written in Java. The audio data was digitized and compressed using the GSM encoding standard. The GSM encoding required only 13 Kbits/second, which meant that even on a 28K line, half the bandwidth was available for the motion data. The audio encoding and playback was handled entirely by a freeware application called Speak Freely. Speak Freely was chosen because it was cost-free and available with full source code in C for both UNIX and Windows. The resulting compressed data packets were encapsulated using RTP (Real-Time Protocol) and sent via UDP transport to users through a network of data relays. Initially there were concerns over whether a Java application might be able to keep up with the steady stream of data, but those fears turned out to be unfounded. Provided the number of listeners accessing each relay was kept to a reasonable number, software performance was quite acceptable.

Elements of the VRML Dream project were first presented in public at the Web3D RoundUP held at the VRML 98 Symposium in Monterey, CA. A modified version using motion capture data was shown at the VRML 99 Symposium Web3D RoundUP in Paderborn, Germany. In both cases, Matsuba and Roehl found the feedback gained from the audience extremely helpful. They not only received constructive criticism, but were also offered help both in manpower and technical resources. [Guest Editorsí note: When difficulties with the audio stream were encountered in mid-performance, their improvisation of human and supernatural character voices got the biggest applause of the night.]

The VRML Dream performance was multicast live over the Internet on April 26, 1998. The production had problems with the transmission of the live audio due mostly to Speak Freelyís inability to handle large multicasts. In addition to the audio problems, there were some viewers who did not receive the streaming motion data. Despite these problems, the VRML Dream project was a success. The motion data system worked for the majority of users, and subsequent tests have shown that the motion and audio streaming technology does work. Moreover, interest in the technology continues to be high. Indeed, the response to the project has been overwhelming. Even though the project limited its press releases largely to specific mailing lists, the website received an average of 600 hits per day in the weeks leading up to the performance.

Based on this success, Matsuba, Roehl and Jim Evans have formed a company called VR Studios in order to continue developing the technology and create new projects. The group is already working on ways to improve the audio quality and has developed interfaces for controlling avatars with motion capture equipment. It is also experimenting with a number of Web3D technologies including Pulse3D, Blaxxun3D and Shout 3D. New entertainment projects currently in the planning stages include an adaptation of Hamlet and a science fiction series called Bots. VR Studios is also developing innovative streaming, real-time 3D applications for e-commerce and corporate training. More information is available at


  1. Coco, Donna. "A Web Developerís Dream," Computer Graphics World 21:8 (August 1998), pp. 54-58.
  2. Matsuba, Stephen N. "Artistic Vision Statement for VRML Dream,"

Stephen Naoyuki Matsuba is a founding partner in VR Studios and is co-chair of the VRML Streaming Working Group, charged with establishing standards for streaming media in VRML. As well as being the co-founder (with Bernie Roehl) of the original VRML Dream project, Matsuba served as its producer, creative director, publicist and project manager. He was also one of the projectís 3D modelers.

Stephen Naoyuki Matsuba
VR Studios

The copyright of articles and images printed remains with the author unless otherwise indicated.