This paper describes the systems developed by Biometric Vox for the Albayzin Speaker Diarization Challenge organized as part of the Iberspeech 2020 conference. The two systems (primary and contrastive) we developed for the challenge are based on Deep Neural Network x–vector embeddings and a PLDA backend. The resulting x-vectors are grouped using Agglomerative Hierarchical Clustering (AHC) in order to obtain the diarization labels. Systems differ in the resegmentation stage. Our primary system achieves 14.96% DER on the test set of the RTVE2018 database and 21.35% on the 2020 evaluation set.