Photo-realistic talking avatars are photo animations of human faces from single photos with pre-recorded sound-tracks and expressions. Its main use is multimedia personal entertainment.
Technology
The technology used by some commercial services is based in creating 3D simulations of human heads and two principles: *the same mesh topology can be morphed to recreate basic human expressions *animations (motion tracks, in fact) can be transferred from mesh to mesh to match movements, expressions and acting gestures of the original acting mesh
The specification of the photographs required are common to virtual face recognition and reconstruction software based in face images: front look, mouth closed, relaxed expression, uniform lighting over all the face extension without shinning areas, over-exposition or hard contrasts of shadows.
Resulting animations convey a true sense of realism and believability but the convenience of using a single photo is also its main limitation; it doesn't model hair or head profiles as 3D scanner hardware stations and systems reverse engineering based on multiple calibrated photographs do.
|