The large development experienced by social networks has unveiled security concerns related to potential attacks on biometric systems. In particular, several examples have shown how a non-authorised subject can easily record the voice a given person and use it to gain access to numerous applications. The most recent ASVSpoof 2019 competition showed that most forms of attacks can be detected reliably with ensemble classifier-based presentation attack detection (PAD) approaches. These, though, depend fundamentally upon the complementarity of systems in the ensemble. In order to improve the generalisability capability of voice PAD solutions, we present a two-stream Convolutional Neural Network that combines the information stemming from two spectrogram-to-image techniques to enhance state-of-the-art baselines.