Understanding the maintenance and origin of biodiversity is a formidable task, yet many ubiquitous ecological patterns are predicted by a surprisingly simple and widely studied neutral model that ignores functional differences between species. However, this model assumes that new species arise instantaneously as singletons and consequently makes unrealistic predictions about species lifetimes, speciation rates and number of rare species. Here, we resolve these anomalies - without compromising any of the original model's existing achievements and retaining computational and analytical tractability - by modelling speciation as a gradual, protracted, process rather than an instantaneous event. Our model also makes new predictions about the diversity of 'incipient' species and rare species in the metacommunity. We show that it is both necessary and straightforward to incorporate protracted speciation in future studies of neutral models, and argue that non-neutral models should also model speciation as a gradual process rather than an instantaneous one.