Introduction - The genetic diversity of wild bananas can conveniently be preserved in seed form. However, depending on the species and experimental sowing conditions, such seeds give uncertain, highly variable germination rates in the greenhouse. One hypothesis would be that the integuments of banana seeds contain dormancy factors preventing good germination when sown naturally. The aim of this work was to ascertain whether the germination rate for banana could be increased if embryos were removed from the influence of the integuments by culturing isolated embryos in vitro. Materials and methods - We present here our results for over 12,000 sowings of a vast panel of Musa acuminata and M. balbisiana genotypes grown in a collection at CARBAP in Cameroon. After harvesting at maturity from self-pollinated banana plants, fresh seeds were immediately sown on two contrasting sterile substrates in a greenhouse at CARBAP and were also cultured in vitro using embryos excised from the same seed batches. Results and discussion - We found substantial variations in germination percentages in the greenhouse depending on the genotypes studied. M. balbisiana seeds germinated well, with an average rate of 80%, or even peaks of 90% for some batches. On the other hand, the different M. acuminata accessions germinated poorly, suggesting dormancy factors in the seeds. We found that in vitro embryo culture eliminated all these differences between the genotypes, with a strong increase in germination rates for the genotypes studied. Conclusion - These results support the hypothesis that dormancy factors in Musa seeds are located in the seed integuments, rather than in the actual embryos. The implications of these results are numerous for collecting and managing batches of Musa seeds, and for conserving Musa genetic resources.