Ongoing global climatic changes and growing demographic pressure have increased demand for agronomic resources and affected the agroecosystem by provoking a number of abiotic stresses that, added to biotic ones, result in physiological and metabolic disorders. Such stresses ultimately impact yield when it most needs to be improved, and understanding and resolving them is a major scientific and agronomic challenge of this century. However, many species are difficult to breed for stress resistance and improved yield for a number of reasons, ranging from a long life cycle (woody species), a reduced genetic background (most self-fertile, cleistogamous legumes) or conversely extensive heterozygosity resulting from an outbreeding nature, and also due to the mainly multigenic origin of such resistances. Biotechnologybased breeding would be an efficient alternative but, for recalcitrant crops, many attempts at in vitro regeneration met with varying degrees of success and often limited to a few genotypes, hampering exploitation of biotechnology approaches. To reduce the risk of undirected somaclonal variations amongst regenerants and transformants, it is better to produce them through somatic embryogenesis that recognises a single-cell origin but whose feasibility is also limited amongst species. There is also a need to fix the resulting genome once a novelty is obtained to ensure efficient heritability of improved traits acquired, which takes several generations in conventional breeding. Acceleration of generations through flowering and fruit set in vitro has been developed in various species including legumes. Haplo-diploidisation in vitro also offers a unique alternative to conventional methods, as it yields novel genetic combinations following doubling of haplotypes and regeneration of fertile plants having gained homozygosity within a single generation. This review will examine the relationships between embryogenesis, stress and its impact on in vitro development of novel genotypes more apt for a sustainable agriculture.