A system and methods are disclosed for serving combined media using resource metadata referencing original media data. Combined resource metadata is created from combining the metadata of two or more files for a virtual media file representing the requested combined media. The combined data can be adjusted to synch audio and video portions as needed for combining the files. The virtual media file references the original media files and serves the content on-the-fly, rather than creating a new file with the combined media, at a savings of server response time and memory consumption, while maintaining high performance.