This paper describes a multi-modal person verification system using speech and frontal face images. We consider two different speaker verification algorithms, a text-independent method using a second-order statistical measure and a text-dependent method based on hidden Markov modelling, as well as a face verification technique using a robust form of corellation. Fusion of the different recognition modules is performed by a Support Vector Machine classifier. Experimental results obtained on the audio-visual database XM2VTS for individual modalities and their combinations show that multimodal systems yield better performances than individual modules for all cases.