Background: Shoulder dystocia (SD) requires emergent intervention to prevent maternal and fetal harm, and simulation models for training can be expensive. We developed a novel, cheap and easily transportable low-fidelity simulation (LFS) model to compare to a commercially available high-fidelity simulation (HFS) model. Methods: Emergency medicine residents were randomized to training on the HFS or novel LFS model. Subjects completed a pretest and a 1-week and 6-month posttest including a self-assessment and a simulated SD delivery. Results: Twenty-seven of the 43 residents completed the study (63%). The number of individuals performing dangerous maneuvers at baseline was similar, 1 week after training was five in HFS and 11 in LFS (p = 0.08) groups and at 6 months was again similar between groups. Mean checklist scores for appropriate actions increased 1 week after training but returned to baseline by 6 months and were similar between groups. The rate of successful delivery, median time to successful delivery, and maximum force applied improved at 1 week and was sustained at 6 months in both groups. Conclusion: Within our limited study population, we did not find a large difference in the occurrence of dangerous actions during simulated SD delivery following HFS and LFS training. Our novel and easily transportable LFS trainer, assembled for less than US$10 each, may be a useful tool to train inexperienced providers on the steps of this procedure. However, this requires further study, as does whether HFS models with force monitoring capabilities may be helpful to train providers to minimize dangerous maneuvers such as the application of excessive force.