In general, in an aspect, versions of representations of segments of a media item are received and temporarily stored on a media playing device. The segments have a sequence. The temporarily stored versions are used in sequence to present at least part of the media item to a user. One or more other versions of representations of segments of the media item are also received and persistently stored on the media playing device. The one or more persistently stored other versions are used to enhance the presentation of at least part of the media item to the user.