I start with a brief overview of photometric redshifts. I describe the new version of the Bayesian Photometric Redshift software, BPZ, and compare its performance with other public photo-z codes as EAZY or LePhare on datasets like COSMOS or ALHAMBRA. I also introduce two large cosmological projects which heavily rely on photometric redshifts: the MultiCycle Treasury "Clusters and Supernova with Hubble" project, and the Javalambre-PAU Astrophysical Survey.