The aim of this study was to identify a gait training type that better improves the walking and balancing abilities of adult patients with chronic hemiplegic stroke. Single-blinded, randomized, controlled, comparative preliminary study was carried out. Patients were recruited from the inpatient unit of a Rehabilitation Hospital. Thirty-one patients who had experienced hemiplegic stroke were randomly assigned to three groups: the heel group (gait training by active weight bearing on the paretic heel with auditory feedback), the forefoot group (gait training with auditory feedback from paretic metatarsals), and the control group (general gait intervention). All patients performed 30 min of comprehensive rehabilitation therapy followed by an additional 20 min of gait intervention with or without auditory feedback three times a week for 6 weeks. Significant improvements in walking and balancing variables were observed after gait training in all three groups (P<0.05). However, significantly larger gains were identified in the heel group than in the control group (center of loading path length, −29.4 vs. −11.4%, d=−1.0; center of loading path velocity, −35.8 vs. −19.6%, d=−1.4). In addition, significantly larger gains were observed in the forefoot group than in the control group (functional gait assessment, +42.6 vs. +20.1%, d=1.3; center of loading path length, −37.2 vs. −11.4%, d=−1.8; center of loading path velocity, −36.0 vs. −19.6%, d=−1.3). Auditory feedback during active weight bearing on paretic feet appears to more effectively improve the walking and balancing abilities of hemiplegic stroke patients than general gait training.