The archive pass
Archive finished video projects.
What to keep, what to throw away, and where the rest goes.
A delivered job is sitting on your working drive. You cannot delete it and you cannot afford to keep it online. This is the pass you run once per finished project: two decisions, one keep list, one delete list, a format that will still open in five years, and a restore you actually test.
Video hosting for filmmakers · 5 GB free · Paid plans from USD 9/month
Updated September 2026
One pass, two questions
An archive is not one thing:
decide which one you are building.
Every bad archive starts the same way, by conflating two jobs that need different contents. One of them is large, expensive and rarely opened. The other is small, cheap and the only one most jobs will ever need. Sort the project into one of the two before you copy a single file.
The re-cut archive
You may have to open the project again: a new cutdown, a re-grade, a version with the sponsor removed, a claim that needs the original frames. It has to contain everything the timeline points at, which is why it is measured in hundreds of gigabytes.
The re-delivery archive
Someone will ask for the finished file. The agency lost it, the client changed staff, a festival wants the exact export. It contains the delivered files and a note saying what they are. On most jobs it is about a tenth of the size of the re-cut archive.
Almost every job deserves a re-delivery archive and almost none of them get one, because the reflex is to zip the whole working folder or to keep nothing. The zip is expensive and nobody trusts it. Keeping nothing means the phone call two years from now ends with you apologising.
The test is a question about the client, not about the footage. Is there a plausible future in which this timeline is opened again? A brand film with a two-year shelf life, a documentary, anything with rights that renew: yes, build the re-cut archive. A weekly social cutdown, a recap video for an event that is over, a job for a client who no longer exists: no, keep the deliverables and the manifest and let the rest go.
The keep list
Seven things go in the folder,
and one file that costs nothing.
Selects rather than every card: on most jobs the takes that made the cut are about a fifth of what was shot. The rest stays on your own disks and RAID or goes, depending on what the contract says.
The manifest is the highest-value item on the list
One text file, written in ten minutes, that says what is in the folder, what was delivered to whom and on what date, which application and version cut it, what the deliverable specifications were, where the music came from and what its licence allows, and what was deliberately not kept. It costs nothing and it is the only part of the archive that still works when the software has moved on three versions.
Write it as plain text, not a document, and put it at the top level of the folder with the job name in the file name. Every other item on the keep list is useless to the person who opens this folder in five years if they cannot tell what they are looking at. That person is usually you, and you will not remember.
The delete list
Most of the folder is worth
throwing away.
Working folders are mostly machine output. None of it is irreplaceable and all of it is expensive to keep, because it is the bulkiest part of the project. Delete it before you measure the archive, not after.
Proxies
Generated from the originals by a button. If a re-cut ever happens, the application makes new ones in an afternoon, in whatever format that year's software prefers.
Render caches and autosaves
Cache folders, optimised media, preview renders, the autosave directory with two hundred timestamped copies of the same timeline. All derived, all disposable.
Duplicate transcodes
The set of files someone made because an application would not read the camera format. The camera originals or the selects are the ones to keep.
The eleven exports nobody used
Every finished job has a folder of near-identical exports. Keep the ones that were actually delivered, named so you can tell, and delete the rest.
Culled takes
The takes you rejected in the first pass. This is the largest single saving in the archive, and the one that needs the contract checked first.
Anything the contract owns
If the agreement says the client owns the rushes, or that you must hold them for a set period, the delete list stops there. Read it before the pass, not during.
The format question
Archive to a codec that will still open,
not to whatever the timeline was in.
An archive is a bet that a file will be readable by software that does not exist yet. That rules out anything exotic, anything tied to one plugin, and anything that only decodes inside the application that wrote it.
The mistake worth naming is archiving the delivery file as the master. A 40 Mbps H.264 export is a fine deliverable and a poor starting point for a re-grade: it is the version of the film with most of the picture information already thrown away. If a re-cut is plausible, the mezzanine master earns its size. If a re-cut is not plausible, you are building the re-delivery archive and the question does not arise.
One quiet cost to plan for: a mezzanine master is large, and large files are the ones that fail to move. Check what your destination will accept before the pass. Google, for example, takes files up to 5 TB but caps uploads at 750 GB per 24 hours, which turns a big archive into a multi-day job.
Where it goes
The cheapest options per terabyte are
the ones where retrieval is a job.
That is the whole trade, and it is worth stating before any price. Cheap archive storage is cheap because nothing is expected to come out of it in a hurry. Expensive archive storage is expensive because it sits next to the thing you are working on.
$20
per TB a month, cold here
Cold storage on Max and Studio, beyond the 1 TB included on Max and the 2 TB included on Studio.
60 days
after a lapse
Vimeo states source files stored on Vimeo are deleted 60 days after a paid plan lapses to Basic, and cannot be recovered.
2 years
over quota
Google states that if you are over your storage quota for two years, it might delete your content across Gmail, Drive and Photos.
Our rates, fetched 9 September 2026 from our own pricing page: cold storage exists on Max and Studio only, at $20 per TB a month beyond the included pool. The bucket column describes how pay-as-you-go object storage is billed rather than any one provider's rate. Every column here describes an archive, so no column streams: the archived copy plays only after it comes back.
An archive that depends on an active subscription is an archive with an expiry date
This is the failure mode nobody plans for, and both of the big consumer surfaces publish it. Vimeo states that if a paid membership lapses to Basic, any source files stored on Vimeo are deleted after 60 days and cannot be recovered, while the HD and SD versions of the videos remain. Vimeo also says it considers accounts not used for a full year inactive and reserves the right to remove them. Google states that if you are over your storage quota for two years, it might delete your content across Gmail, Drive and Photos.
None of that is a scandal. It is what those products are: places where files live while you are paying attention. An archive is the copy that survives your attention lapsing, which is why it wants at least two homes, and why the manifest matters more than any single one of them.
Here is the honest position on our own side. Cold storage exists on Max and Studio only: Max is $75 a month or $750 a year for 1.5 TB hot plus 1 TB cold, Studio is $250 or $2,500 for 2 TB hot plus 2 TB cold, and extra seats are $25 a month, or $300 a year, the same on every paid plan, with no storage attached to a seat. Cold blocks beyond the included pool are $20 per TB a month, and a seat does not change that pool. Archiving frees the hot space immediately, restores are size-verified, and the re-hydration allowance is twice your cold pool per calendar month, resetting on the 1st. Masters up to 20 GB move through the web dashboard; larger ones move through the macOS app part by part. Uploads take up to 5 TB per file, subject to your available storage.
We are more expensive per terabyte than a raw bucket you run yourself, and that is not an accident to explain away. What the difference buys is that the archive sits next to the film that was delivered, in the same library, behind one button, instead of in a separate system with its own credentials, its own console and its own bill. If you are comfortable writing lifecycle policies and running a restore job, use a bucket. You will pay less and you will not be worse off. If the words “lifecycle policy” are the reason your last three finished jobs are still sitting on a working drive, the button is the product.
The limitation belongs on this page in plain words, because it is the one that surprises people: a cold master does not stream, does not embed and does not open in the review room. It is a stored file, nothing else, until you bring it back. Keep the viewing export hot if anyone might click a link, and send the master to cold on its own.
Not a technical question
The one decision this page will not make for you is
how long you keep it.
Retention is a business decision wearing a storage costume. Three things settle it and none of them is a codec. What does the contract say, in the clause about materials and ownership, about how long you hold the working files and who owns the rushes. What is the realistic window in which this client comes back, which is usually shorter for social work and much longer for anything with a licence that renews. And who pays for the years in between: some studios bill an archive line, some fold it into the project fee, and some absorb it silently until the storage bill becomes a surprise.
Decide it once, per client type, write the answer into your contract template, and put the retention date in the manifest of every job you archive. A folder that says when it may be deleted is a folder someone can delete. A folder that says nothing lives forever, which is the most expensive outcome available.
The part everyone skips
An archive nobody has ever restored from
is a hope, not an archive.
Pick one job a year, at random
Not the one you remember packing carefully. Random selection is the only kind that tests the habit rather than the exception. Put it in the calendar on a fixed date so it survives a busy quarter.
Restore the whole folder, not a file
Bring back everything the manifest says is there, to a scratch disk, and check the sizes against the manifest. A restore that only pulls one file proves the storage works and proves nothing about the archive.
Open the project and relink
This is where archives fail: missing fonts, a plugin that no longer installs, a graphics project that opens with placeholder text, a LUT that was referenced from a path that no longer exists. None of that shows up until someone opens the timeline.
Write down what broke, then fix the template
Every failure you find is a line to add to the keep list for future jobs. The point of the yearly restore is not this project. It is that the next fifty archives do not have the same hole in them.
If you want the shortest possible version of this whole page: keep the deliverables and a manifest for every job, keep the mezzanine master and the project for the jobs that might be re-cut, delete the machine output, put the result somewhere that does not depend on a subscription you might forget, and open one of them a year to prove it works. The rest is arithmetic, and the hot versus cold breakdown has the numbers for a full year of work.
Questions
Frequently asked
What is the smallest archive that is still worth keeping?
The delivered files plus a plain text manifest. That is the re-delivery archive: it answers the call that actually comes, which is a client asking for the film again two years later. It is roughly a tenth of the size of a re-cut archive, and on most jobs it is the only archive you will ever open.
Should I archive the ProRes master or the H.264 deliverable?
Both, for different reasons. The mezzanine master (ProRes 422 HQ, ProRes 4444 or DNxHR) is the file a re-cut or a re-grade starts from, and it is worth its size. The delivered H.264 or HEVC file is what the client asks for by name, and it is small. Archiving only the deliverable saves space and quietly ends the possibility of a re-cut at full quality.
Can I just leave everything on my video host?
Only while you keep paying, and only if you read the terms. Vimeo states that source files stored on Vimeo are deleted 60 days after a paid membership lapses to Basic, and cannot be recovered, while the HD and SD versions remain. Vimeo also says it reserves the right to remove accounts not used for a full year. Google states that if you are over your storage quota for two years it might delete your content across Gmail, Drive and Photos. A hosting account is a delivery surface. An archive is a copy you control.
How big does a video archive have to be before the cloud beats a shelf of drives?
It is not a size question, it is a habit question. A drive on a shelf is the cheapest copy you can make and the easiest one to forget, mislabel, drop or leave unspun for four years. If you have a labelling and rotation habit that you actually follow, keep using it and add one off-site copy. If you do not, pay for the off-site copy first and treat the drives as the second copy.
What does it cost to keep an archive in cold cloud storage?
A bucket you run yourself is the cheap floor: pay-as-you-go cold object storage bills by the gigabyte you keep, with no tier to buy, and bills the gigabytes you pull back as egress. Here, cold is $20 per TB a month, with 1 TB included on Max and 2 TB on Studio. We are more expensive per terabyte, on purpose, and the section above says what that buys.
How do I get a 400 GB master into cold storage here?
The web dashboard moves masters up to 20 GB. Anything larger moves through the macOS app, which sends the file part by part, so a laptop that closes mid-transfer resumes instead of restarting. Uploads take up to 5 TB per file, subject to your available storage.
How often should I test a restore?
Once a year, on one real job, chosen at random rather than the one you remember packing well. Restore it, open the project file, relink, and write down everything that did not come back. An archive nobody has ever restored from is a hope, not an archive.
Archive beside the film, not in another system
Cold storage lives on Max and Studio: 1.5 TB hot plus 1 TB cold on Max, 2 TB hot plus 2 TB cold on Studio, extra cold at $20 per TB a month. Archiving frees the hot space the moment it runs.
Sources
- 1.Vimeo Help Center: FAQ, lapsing from a paid account to a Basic account (source files deleted after 60 days and not recoverable; HD and SD versions remain; fetched 9 September 2026)
- 2.Vimeo Help Center: About the Vimeo Free plan (1 GB for the account lifetime; accounts unused for a full year may be removed; fetched 9 September 2026)
- 3.Vimeo Help Center: how storage is calculated (storage consumed is the size of the original source file at upload; fetched 9 September 2026)
- 4.Google Drive Help: what happens when you reach your storage limit (over quota for 2 years, Google might delete content across Gmail, Drive and Photos; fetched 9 September 2026)
- 5.Google Workspace: storage and upload limits (750 GB upload per 24 hours, individual files up to 5 TB; fetched 9 September 2026)
- 6.uncompressed.io pricing: plans, hot and cold storage blocks
- 7.uncompressed.io: the Vault, and the storage and seats each plan includes