The nature of multi-featured object representations in visual working memory remains unclear. Here, we aim to test two competing hypotheses. One is that all object features are integrated into an object file. The other is that locations are critical for maintaining the binding of visual features of an object in VWM. We developed a hybrid paradigm that bridges the VWM and object-file literatures. A set of multi-featured objects were presented simultaneously, followed by a retention interval in which placeholders either moved or stayed. Participants were then given one feature as a cue and prompted to report the other two features. Applying MPT models to evaluate these competing accounts, we found that the data are better explained by the object file framework rather than the spatial binding hypothesis. Our findings suggest that maintenance of bindings between nonspatial features, and of these features with locations, are possible regardless of the objects’ motion.