Embodiments of the present disclosure provide techniques for rendering content from a media item. According to these embodiments, from a file of the media item, track(s) in a group data structure corresponding to the type of content are identified as candidate track(s). From other tracks in the file, a determination may be made whether another track corresponds to the type of content. When another track corresponds to the type of content, feature tags in the file that are associated may be compared with the other track to capabilities of a player device that is to render the type of content. When the feature tags match capabilities of the player device, the other track may be included as a candidate track. Thereafter, a track may be selected from the candidate tracks and rendered by the player device.