Despite advances in screening technologies, only a very small fraction of theoretical protein sequence can be sampled in directed evolution experiments. At the current state of random mutagenesis technologies mutation frequencies have often been adjusted to values that cause a limited number of amino acid changes (often one to four amino acid changes per protein). For harvesting the power of directed evolution algorithms it is therefore important that generated mutant libraries are rich in diversity and enriched in active population. Insufficient knowledge about protein traits, mutational robustness of protein folds and technological limitations in diversity generating methods are main challenges for managing the complexity of protein sequence space. This review covers computational and experimental advances for high quality mutant library generation that have been achieved in the past two years.