We present recent developments on our system of visual articulatory feedback, named Vizartd3D, which is based on the 3D orofacial clone developed at GIPSA-lab. This system aims at providing any speaker with a feedback on his/her own articulation, especially by displaying the tongue movements. Application areas are computer-assisted pronunciation training (phonetic correction) for second-language learning and speech rehabilitation. In our system, the orofacial clone is animated in real-time from the speech audio signal, using a statistical mapping method based on multiple Gaussian mixture regressions (GMR). This approach combines in the same framework a spectral mapping step with an acousticarticulatory inversion step. All the modules of the system are implemented in the real-time environment Max/MSP (acoustic analysis, mapping and 3D rendering). More information is available at http://www.gipsa-lab.fr/projet/vizart3D/.