Archival and Master File Management

What a master file should be, how to store it so it is still usable in ten years, and the verification step that makes an archive trustworthy.

A master is the file everything else is derived from, and a surprising number of projects do not have one. The finished film exists as a compressed delivery file, the project has been closed, and the next time a version is needed the only available source is something that was already encoded for distribution. Re-encoding from that compounds the compression, and the second generation is visibly worse than the first.

The master should be encoded in a mastering codec rather than a delivery codec: every frame complete and independent, minimal compression, full colour information, and the audio at full quality. It is large, and that is the point. It exists so that every future version, in every format, at every resolution, can be produced from a source that has not already been degraded.

There should also be a textless master, which is the film without titles, lower thirds, subtitles or endframe text. This is what makes future language versions and rebranded versions possible without rebuilding, and it is the single most valuable archival item for a company that operates in more than one market. Producing it at delivery costs one additional export; producing it two years later is not possible at all.

Stems for the audio serve the same function. Dialogue, music, effects and ambience kept as separate tracks mean a destination specific mix, a different language voiceover or a silent version can be produced by rebalancing rather than rebuilding. Zhang et al. (2025) found that audiovisual features contribute measurably to engagement, and the ability to produce a properly balanced mix per destination later depends entirely on whether the stems were kept.

The archive around the master is what determines whether it is usable. Source media, project files, the fonts used, the licensed music with its licence documents, the 3D scenes, the generated material with the settings that produced it, the approved brand assets, and a readme naming what is what. A master without its project is a film that can be played and not changed, which is half of what an archive is for.

Storage should follow a simple rule that most studios violate: a single copy is not a backup. The workable arrangement is at least two copies on separate physical devices, with one of them off site or in cloud storage. Drives fail, offices flood, and the probability of losing a project is not low over a ten year horizon. The cost of a second copy is trivial against the cost of reshooting.

Verification is the step that separates an archive from a folder of files that are assumed to be fine. Generating a checksum for each file at the time of archiving, and re-verifying periodically, detects silent corruption, which does occur and which is otherwise discovered only when a file is needed. An archive that has never been verified is an archive whose condition is unknown.

Format obsolescence is a slower risk and a real one. Codecs, container formats and project file versions all age, and a project file from a decade old version of an application may not open in a current one. The mitigation is to keep the master in a widely supported mastering format rather than an exotic one, and to migrate the archive when a format is clearly declining rather than when it has already stopped working.

Retention should be a stated policy rather than an accumulation. Keeping source media indefinitely is expensive and rarely necessary; keeping masters, textless masters, stems, project files and generative records indefinitely is cheap and almost always worthwhile. A policy that keeps source media for a defined period and everything else permanently gives most of the benefit at a fraction of the storage.

The organisational point that makes all of this work is that archiving is a delivery task rather than a tidy up performed when someone has time. It happens at the end of the project, while the material is still organised and the person who made it still remembers what is what. Archives assembled six months later are incomplete, because the knowledge required to assemble them has already dispersed.

References

Zhang, Z., Qiu, K., & Ye, Y. (2025). Influence of audiovisual features of short video advertising on consumer engagement behaviors: Evidence from TikTok. Journal of Business Research, 201, Article 115662. https://doi.org/10.1016/j.jbusres.2025.115662