Thirty-nine adult individuals with stroke in the stable phase were asked to walk at their preferred speed and then as fast as possible. A set of gait indexes were computed, including spatial temporal parameters, ankle and hip mechanical work, and timing of ankle push-off onset, for comparison with normal velocity-dependent profiles. Algorithms were used to classify the resulting gait patterns when the patients walked at their preferred speed and fast and to identify the patients' strategies to maximise speed. Patients' strategies were characterised by a variation in the parameters, which were reduced, equal or increased, in relation to normal patterns. At both speeds, stroke individuals tended to walk at higher cadence and with shorter stride length. At the preferred speed the investigated parameters for all patients were mostly within the normal profile (71.8-94.9%). The exception was the finding of positive work at the ankle where 64% of the stroke individuals showed reduced work production. At fast speed (increments to 36%BH/s) fewer patients presented values within the normal profile for all the parameters (17.9-74.4%), with the exception of negative work at the ankle and hip. The parameter variations showed a more consistently abnormal picture. The results indicate that, in order to increase gait speed, patients with hemiparesis have different functional resources on which to draw, and these vary from individual to individual. Thus, gait analysis at different gait speed should be adopted to develop individualised programs that will improve quality of life for the patients.