Multilingual Automated digital talking character
Kostas Avdelidis, Charalampos A. Dimoulas, C. Bliatsiou, T. Pasias, Georgios Kalliris, Γεώργιος Παπανικολάου, Giannis Stoitsis · 2002
The current paper focuses on the development of a 3D character generation and multilingual talking emulation software application. Traditional lip sync techniques in 3D graphics productions require a careful (and painful) design at the pre-render phase, as well as the selection of the mostly matching phrases during audio dubbing. This paper proposes a completely different approach, whereas graphic designers are not worried about facial animation for lip sync purposes. When the scene’s sequence modeling is finished, rendering starts just after an automatic modification procedure where faces are morphed according to the driving sound. In case of a multilingual production, rendering for each different language is required, but the content of the dialogues is not depended on the original facial expressions formed according to the original language. Based on our experimentation, we suggest a “standardized face ” format for 3d graphic popular platforms in combination with the developed algorithm for automated lip sync, translation and rendering.