Background: Action observation improves excitability of the primary motor cortex and the encoding of motor engrams as well as motor-learning. Objective: The intention of our pilot-study was to evaluate the feasibility of a six weeks home-based action observation training (video therapy) in stroke patients. Methods: 56 patients (age 58 ± 13; time since onset 40 ± 82 months; NIHSS 3.5 ± 1.8) with a hand paresis following stroke were recruited from two rehabilitation clinics. Before discharge from the clinic the intervention group received a DVD displaying ten object-related motor tasks of varying difficulty, each lasting five minutes. Patients were requested to imitate the motor tasks one hour daily for six weeks (“video group”). A control group performed the same tasks with written instructions without observation/imitation (“text group”). A second control group was discharged without specific homework (“usual care group”). Results: There was no dropout in the video group. Quality and speed of the Motor Activity Log (MAL) increased significantly in the video and text group. Nine Hole Peg test (NHPT) and Stroke Impact Scale (SIS) improved only in the video group. Questionnaires (MAL and SIS), obtained twelve months after training in fourteen and eleven participants of both active groups, indicated significant differences in favor of the video group. Conclusions: Video training is easy to deliver and highly accepted by patients. Six weeks of home based training suggests improvement of hand function, activities of daily living and quality of life. Video-therapy appears to be promising, as an adjunct to conventional neurorehabilitation – especially with regards to non-supervised, homebased training.