Each year, health-care providers face the challenge of providing a timely and proportionate response to the annual seasonal influenza epidemics. Accurate forecasting of these epidemics can enable these health-care providers to maximize preparedness. For several years now, we have used a mechanistic transmission model coupled with a statistical observation model to generate predictions of key epidemic properties such as peak time and magnitude. Starting in Melbourne, we now generate forecasts for several capital cities around Australia using a particle filter, the state of the art in Bayesian forecasting. In addition to the practical outcome of improving our forecasting capability, this work emphasizes a key strength of the Bayesian paradigm, the capability to seamlessly incorporate prior knowledge. While the issue of “rigged-priors” has made this issue somewhat controversial, we feel the method provides an honest and beneficial approach and will enable more nuanced modelling