The Timed Up-and-Go (TUG) test has been combined with different verbal/cognitive tasks (i.e., TUG dual task [TUGdt]) as a form of motor-cognitive testing. However, it is still unclear how different TUGdt conditions affect gait among older adults. Thirty community-dwelling older adults, with mean age of 73 years, participated in the study. Data were collected using marker-free video recordings. Gait parameters were extracted using a semiautomatic deep learning system. Comparisons of execution time and gait parameter outcomes were made under TUG and three types of TUGdt test conditions: TUGdt-naming animals, TUGdt-months backwards, and TUGdt-serial 7s. Statistical analyses were based on mean values of the gait parameters for each participant and TUG condition, including TUGdt gait cost, that is, the relative difference between TUGdt and TUG. All the investigated TUGdt conditions resulted in varying degrees of gait parameter changes. Under TUGdt conditions, participants took shorter and slower steps, with TUGdt-serial 7s causing the largest interference.