FaceSpace

Photorealistic Speech Agents

A front end for intelligent systems including the ProtoAndroid and VR based theater

facespace image medely

a) FaceSpace in 1991 was a cartoon character exploring facial and spoken expression. b) This screen grab shows FaceSpace 2.0, which supports multiple social agents with billions of discreet states, yet is fast and compact. c) Runtime video grab of a current production of "ROMeo and Julie@" (Web Side Story). d) FaceSpace is aimed at advanced interfaces for next generation systems.


FaceSpace interfaces with an Emotion Engine which computes realistic emotional behavior from user defined inputs. Want to build your own Face Space Avatar? Take a look at a sample script and the current Feelings List.

face parts

FaceSpace is a finite state machine begun in FoxPro 2.1, a relational database which we've abused as a media object base. A memory leak which would normally never affect a vanilla FoxPro session causes FaceSpace to crash within minutes. The development team is upgrading to FoxPro 3.0 which is supposed to cure this class of problems.

Input to FaceSpace takes the form of nearly natural language scripts which are parsed for affective content such as emotions, facial states, and speech. These scripts can be generated by AI programs or human authors, even children. During 1995, by lucky accident, an e-mail flame was run through FaceSpace and it was uncanny to watch the facial agent "read" the message. The face took an emotional rollercoaster ride and occaisionally spoke out key words. EmotionSpace is a FaceSpace compatible generator of freerunning mood and emotion.


Ancient FaceSpace

early facial state space

This is the 1989 state map of Face Space 0.1, an eightyone state alphabet of facial expression operated upon by matrix algebra. It was used for concept study to establish principles extended to the much larger state machines that followed.


FaceSpace Creators and Supporters

The FaceSpace development team includes FoxPro/DBase programmers Tim O'Leary and Gilbert Andrade, video specialist John Witham, musicians/audio engineers Fred Mitchum and Courtney Audain, facilitator Mike Wren, and David Santos as schema architect and coordinator. Faustex Systems Corporation provides major support for this work. The Robot Group, with funding from the City of Austin, has sponsored FaceSpace at various Robofests and in outreach to Austin area priority schools.

The Computers, Robotics, and Artists Society of Houston (CRASH) is taking a major role in the development of Web based and 3D FaceSpace.


3D Avatar Animation of Julie@ (168K)

More Details


Thanks for flying... PolyCosmos.


Comment to santos@88net.net


FaceSpace has been sponsored in part by the City of Austin under the auspices of the Austin Arts Commission.