The controlled mixing of streams with different salinity is a potential route for clean and renewable base-load power generation. Here, a comprehensive process model has been developed for pressure-retarded osmosis (PRO) accounting for full-scale system losses such as viscous dissipation, external mass transfer and equipment efficiency. Also, an existing model for reverse electro-dialysis (RED) is adapted to account for analogous full-scale system losses. The models are used to predict practical power densities and process efficiencies. The projected power density for PRO (using best available membranes) is much lower than generally predicted by extrapolation of experimental data For example, a power density of 4 W/m(2) extrapolated from laboratory experiments actually yielded negative power at full-scale. The maximum power density for PRO is doubled as the hydraulic energy recovery (HER) efficiency is increased from 90% to 99%. Furthermore, the operating pressure, load voltage, and crossflow velocities typically applied in laboratory studies appear much too high to be practical in full-scale PRO and RED systems. Notably, RED systems exhibit a lower system size required for achieving a given degree of mixing, compared with PRO. For both processes, maximum energy efficiency does not occur at thermodynamic equilibrium due to hydraulic losses. Finally, maximum power density appears to be an inadequate parameter for assessing full-scale PRO/RED process feasibility because both processes could produce the same maximum power density, yet exhibit different power outputs and efficiencies and system sizes. (C) 2014 Elsevier B.V. All rights reserved.