Background: A velocity offset error in phase contrast cardiovascular magnetic resonance (CMR) imaging is a known problem in clinical assessment of flow volumes in vessels around the heart. Earlier studies have shown that this offset error is clinically relevant over different systems, and cannot be removed by protocol optimization. Correction methods using phantom measurements are time consuming, and assume reproducibility of the offsets which is not the case for all systems. An alternative previously published solution is to correct the in-vivo data in post-processing, interpolating the velocity offset from stationary tissue within the field-of-view. This study aims to validate this interpolation-based offset correction in-vivo in a multi-vendor, multi-center setup. Methods: Data from six 1.5 T CMR systems were evaluated, with two systems from each of the three main vendors. At each system aortic and main pulmonary artery 2D flow studies were acquired during routine clinical or research examinations, with an additional phantom measurement using identical acquisition parameters. To verify the phantom acquisition, a region-of-interest (ROI) at stationary tissue in the thorax wall was placed and compared between in-vivo and phantom measurements. Interpolation-based offset correction was performed on the in-vivo data, after manually excluding regions of spatial wraparound. Correction performance of different spatial orders of interpolation planes was evaluated. Results: A total of 126 flow measurements in 82 subjects were included. At the thorax wall the agreement between in-vivo and phantom was - 0.2 ± 0.6 cm/s. Twenty-eight studies were excluded because of a difference at the thorax wall exceeding 0.6 cm/s from the phantom scan, leaving 98. Before correction, the offset at the vessel as assessed in the phantom was - 0.4 ± 1.5 cm/s, which resulted in a - 5 ± 16% error in cardiac output. The optimal order of the interpolation correction plane was 1st order, except for one system at which a 2nd order plane was required. Application of the interpolation-based correction revealed a remaining offset velocity of 0.1 ± 0.5 cm/s and 0 ± 5% error in cardiac output. Conclusions: This study shows that interpolation-based offset correction reduces the offset with comparable efficacy as phantom measurement phase offset correction, without the time penalty imposed by phantom scans. Trial registration: The study was registered in The Netherlands National Trial Register (NTR) under TC 4865. Registered 19 September 2014. Retrospectively registered.