Current developments in tissue engineering strategies for articular cartilage regeneration focus on the design of supportive three-dimensional scaffolds and their use in combination with cells from different sources. The challenge of translating initial successes in small laboratory animals into the clinics involves pilot studies in large animal models, where safety and efficacy should be investigated during prolonged follow-up periods. Here we present, in a single study, the long-term (up to 1 year) effect of biocompatible porous scaffolds non-seeded and seeded with fresh ex vivo expanded autologous progenitor cells that were derived from three different cell sources [cartilage, fat and bone marrow (BM)] in order to evaluate their advantages as cartilage resurfacing agents. An ovine model of critical size osteochondral focal defect was used and the test items were implanted arthroscopically into the knees. Evidence of regeneration of hyaline quality tissue was observed at 6 and 12 months post-treatment with variable success depending on the cell source. Cartilage and BM-derived mesenchymal stromal cells (MSC), but not those derived from fat, resulted in the best quality of new cartilage, as judged qualitatively by magnetic resonance imaging and macroscopic assessment, and by histological quantitative scores. Given the limitations in sourcing cartilage tissue and the risk of donor site morbidity, BM emerges as a preferential source of MSC for novel cartilage resurfacing therapies of osteochondral defects using copolymeric poly-D,L-lactide-co-glycolide scaffolds.