Our objective was to determine the effect of addition of dipotassium phosphate (DKP) at 3 different thermal treatments on color, viscosity, and sensory properties of 7.5% milk protein-based beverages during 15 d of storage at 4°C. Micellar casein concentrate (MCC) and milk protein concentrate (MPC) containing about 7.5% protein were produced from pasteurized skim milk using a 3×, 3-stage ceramic microfiltration process and a 3×, 3-stage polymeric ultrafiltration membrane process, respectively. The MCC and MPC were each split into 6 batches, based on thermal process and addition of DKP. The 6 batches were no postfiltration heat treatment with added DKP (0.15%), no postfiltration heat without added DKP (0%), postfiltration high-temperature, short time (HTST) with DKP, postfiltration HTST without DKP, postfiltration direct steam injection with DKP, and postfiltration direct steam injection without DKP. The 6 MCC milk-based beverages and the 6 MPC milk-based beverages were stored at 4°C. Viscosity, color, and sensory properties were determined over 15 d of refrigerated storage. MCC- and MPC-based beverages at 7.5% protein with and without 0.15% added dipotassium phosphate were successfully run through an HTST and direct steam injection thermal process. The 7.5% protein MCC-based beverage contained a higher calcium and phosphorus content (2,425 and 1,583 mg/L, respectively) than the 7.5% protein MPC-based beverages (2,141 and 1,338 mg/L, respectively). Pasteurization (HTST) had very little effect on beverage particle size distribution, whereas direct steam injection thermal processing produced protein aggregates with medians in the range of 10 and 175 μm for MPC beverages. A population of casein micelles at about 0.15 μm was found in both MCC- and MPC-based beverages. Larger particles in the 175-μm range were not detected in the MCC beverages. In general, the apparent viscosity (AV) of MCC beverages was higher than MPC beverages. Added DKP increased the AV of both MCC- and MPC-based beverages, while increasing heat treatment decreased AV. The AV of beverages with DKP increased during 15 d of 4°C of storage for both MCC and MPC, whereas there was very little change in AV during storage without DKP and a similar effect was observed for sensory viscosity scores. The L value of beverages was higher with higher heat treatment, but DKP addition decreased L value and sensory opacity greatly. Sulfur-eggy flavors were detected in MPC beverages, but not MCC-based beverages.