A predictive but tunable model for the liquid viscosity of characterized crude oils was developed by use of the generalized Walther correlation (Walther 1931; Yarranton et al. 2013). The crude oils are each characterized into maltene pseudocomponents and a single C5-asphaltene component. The viscosity model requires two pseudocomponent parameters (A and B), two whole-oil parameters (δ1 and δ2), and binary-interaction parameters. The asphaltene parameters were determined experimentally and fixed for all cases. Correlations were developed for the maltene-pseudocomponent parameters and the binary-interaction parameters. The required data are the absolute temperature, pressure, the C5-asphaltene content, the specific gravity (SG) and molecular weight (MW) of the oil, and the boiling-point distribution of the maltenes. The SG and MW distributions of the maltenes are also required, but are generated from existing correlations. The proposed model predicted the viscosity of five western Canadian and two Colombian bitumens; three American, one Mexican, one Venezuelan, and one European heavy oil; and also a conventional oil from the Middle East with an overall average deviation of 57%. Tuning to a single viscosity data point with a single tuning parameter reduced the overall deviation to 8%, close to the 4% deviation obtained when the model was fitted to the data.