Background: Residual walking deficits are common in community-dwelling people after stroke. Aims: The aim of this study was to determine if a four-month treadmill and overground walking program is more effective than a two-month program, compared with control, at improving walking in community-dwelling people with stroke who walk slowly. Method: A three-arm randomized trial with concealed allocation, assessor blinding, and intention-to-treat analysis involving 102 people with stroke living in the community who walked slowly was undertaken. Experimental group 1 undertook 30min of treadmill and overground walking thrice per week for four-months, experimental group 2 undertook training for two-months, while the control group had no intervention. The primary outcome was walking measured as the distance covered during the six-min walk test. Other outcomes were walking speed, step length and cadence, health status, community participation, self-efficacy and falls. Results: By two-months, the experimental groups, who were both undergoing training, had improved their six-min walk distance compared with the control group. The four-month training group continued training beyond two-months and improved further so that by four months they walked 38m (95% confidence interval 15-60) more than the control group and 29m (95% confidence interval 4-53) more than the two-month training group. However, by 12 months, well after the cessation of training, both experimental groups had returned to near baseline levels, and there was no difference between the groups. Conclusion: Four months of treadmill training results in better walking. However, these effects disappear once training ceases. Therefore, training should be ongoing.