Recent studies have sought to use Microsoft Kinect sensors to measure water surface shape in steady flows or transient flow processes. They have typically employed a white colourant, usually titanium dioxide (TiO2), in order to make the surface opaque and visible to the infrared-based sensors. However, the ability of Kinect Version 1 (KV1) and Kinect Version 2 (KV2) sensors to measure the deformation of ostensibly smooth reflective surfaces has never been compared, with most previous studies using a V1 sensor with no justification. Furthermore, the TiO2 has so far been used liberally and indeterminately, with no consideration as to the type of TiO2 to use, the optimal proportion to use or the effect it may have on the very fluid properties being measured. This paper examines the use of anatase TiO2 with two generations of the Microsoft Kinect sensor. Assessing their performance for an ideal flat surface, it is shown that surface data obtained using the V2 sensor is substantially more reliable. Further, the minimum quantity of colourant to enable reliable surface recognition is discovered (0.01% by mass). A stability test shows that the colourant has a strong tendency to settle over time, meaning the fluid must remain well mixed, having serious implications for studies with low Reynolds number or transient processes such as dam breaks. Furthermore, the effect of TiO2 concentration on fluid properties is examined. It is shown that previous studies using concentrations in excess of 1% may have significantly affected the viscosity and surface tension, and thus the surface behaviour being measured. It is therefore recommended that future studies employ the V2 sensor with an anatase TiO2 concentration of 0.01%, and that the effects of TiO2 on the fluid properties are properly quantified before any TiO2-Kinect-derived dataset can be of practical use, for example, in validation of numerical models or in physical models of hydrodynamic processes.