Client re-delivery
Round two of notes always ends the same way: a new export, and a client sitting on a link that points at the old one. Replacing the file behind that link keeps the URL, the embeds, the view count, the password and every setting on the film. Here is exactly what the swap changes, what it destroys, and when you should not use it.
No credit card required. Cancel any time.
Updated September 2026
The mechanism
In the dashboard the control is called Upload a new version. It writes the new master onto the record the client is already pointing at, instead of creating a second one beside it.
The distinction matters because almost everything a delivery depends on is attached to the record, not to the file. The stored file is swapped, and the numbers that describe it are rewritten from the new upload: size, duration, frame rate and dimensions. The record keeps its id and its slug, so the watch URL does not move. It keeps its embed options, so every iframe already placed keeps working. It keeps its view count, its visibility and its password. The old object is then deleted from storage.
That is the whole feature. There is no second link to send, no note to write explaining which of two links is current, and no page anywhere that quietly keeps serving last week’s cut. A client who forwarded the link to three colleagues has, without doing anything, forwarded the new cut to three colleagues.
Behaviour verified in the uncompressed.io codebase on 5 September 2026. The dashboard control is labelled Upload a new version.
The round trip
Round one lives at the link you already sent. Timecoded comments and approvals sit against the version the client watched, so nobody is describing a moment at a timecode that no longer exists.
Export the new cut the way you exported the last one, then archive the previous export on your own drives before you touch the dashboard. The replace deletes the old file here, so your local copy is the only copy of it that will exist.
Open the film, choose Upload a new version and drop the file. The transfer moves in 10 MB parts with up to 60 attempts per part and adaptive lanes, so a hotel connection recovers instead of failing. It keeps running while you move around the app.
There is no new address to send. The message is one line: the new cut is at the same link. The client presses play on the URL they already have and sees the file you just uploaded.
50 GB
Per file
The ceiling for any single upload. The check runs in the browser.
10 MB
Part size
The file moves in parts, so a stall costs one part, not the transfer.
60
Attempts per part
A part keeps retrying through a long dead patch before it gives up.
0
Processing steps
Nothing is transcoded, so the new cut is live when the upload finishes.
What the resume actually does
If a transfer is interrupted, progress is remembered and a resume banner lists the interrupted file. It does not mean the upload survives a reload or a closed laptop: the browser has discarded the bytes. You re-drop the same file, and it continues from the last finished part rather than starting again. The upload does keep running while you navigate around the app.
The part nobody plans for
On a platform that re-encodes, uploading is only the first half of the job. The second half is a queue, and until it drains the version your client is judging is not the version you exported.
Google documents this plainly for YouTube. Its help pages state that “YouTube always re-encodes videos to optimize their playback quality”, that “when you upload a video, it will initially be processed in low quality”, and that a 60-minute 4K 30 fps video “can take up to 4 hours to finish high-resolution processing”. Vimeo’s help centre describes the same shape of step: during transcoding “your video file is re-encoded into several formats that will be available to view on Vimeo”, and its published playback ceilings top out at 22 Mbps for 4K and 7 Mbps at 1080p60.
For a re-delivery, that window is the dangerous part. A client who opens the link ten minutes after you say the new cut is up is not watching a soft version because your grade is soft. They are watching an intermediate. The notes that come back from that viewing are notes about an encode, and you will spend the next call explaining a problem that does not exist in your file.
Nothing is transcoded here and there is no rendition ladder, so there is no intermediate to catch. The player streams the exact bytes you uploaded, which means the new cut is live the moment the upload finishes and the first frame anyone sees is yours. The honest trade is on the other side of that: with no lower rendition to fall back to, a thin connection buffers instead of quietly switching to a mushy version. For an approval round, buffering that resolves is better feedback than a soft picture that does not.

Competitor figures and quotations from Vimeo and Google help centre pages, fetched 5 September 2026 and listed in the sources below.
Be honest about it
The old object is deleted. That is the cost of keeping one link clean, and it is a real cost worth stating before you rely on it.
Once the new file lands, the previous one is gone from storage here. Archive every export you deliver on your own drives, the same way you would keep a project file. If you cannot afford to lose v1, do not swap it.
A replace leaves you with the current cut and nothing else. Nobody can open v1 and v2 side by side afterwards, because v1 no longer exists at that address or any other.
A team that wants to compare cuts, keep per-version approvals and see who signed off on which one is describing a review tool, not a replace button. That is what the review room is for, and it is the right place for the rounds before the swap.
Everything that could open the film yesterday can open the new cut today. That is the point of the feature, and it is also its sharp edge: replacing a file is not a way to take a delivery back from someone.
The clean division is worth spelling out. Rounds of notes belong in the review room, where versions sit beside each other and approvals are recorded against the cut they were given for. The replace is for the moment after that: the film is signed off, the link is already in the client’s hands, in their deck and possibly on their site, and the only thing that should change is the file behind it.
If access is the thing you actually need to change, replacing the file does not do it. A swap leaves visibility and the password exactly as they were, which is deliberate. Cutting people off is its own operation, covered in revoking access to a shared link.
Which move to make
Most re-deliveries are the left column. A colour fix, a corrected supered name, a logo the client changed after sign-off: none of those are new work in the client’s mind, and handing back a new URL for them makes a two-minute fix look like a new deliverable. The right column is for cuts that are genuinely separate objects, where somebody will eventually need to open both.
One practical note for anyone arriving from another platform. If the copy of a film you are hosting today came across as whatever the old platform served rather than your master, the replace is also the upgrade path: put the true master onto the same record later, and every link and embed you have already placed starts serving it. The wider handoff, including which files a finished delivery should contain, is covered in delivering final video files to clients.
Questions
No. The replace writes the new file onto the same record, so the id and the slug are untouched and the watch URL is character for character the one you already sent. A client who bookmarked it, forwarded it or pasted it into a thread gets the new cut at that address the next time they press play.
No. The embed options and the film id are untouched by a replace, so an iframe on a client's site, in a deck or inside an internal wiki keeps playing and starts serving the new file. You do not have to hunt down every page that carries the film to hand over a revision.
No. Views are stored on the record, and the replace only rewrites the file and the numbers that describe it: size, duration, frame rate and dimensions. The count carries across the swap, which is the honest behaviour for a delivery and the reason a replaced film keeps its history.
No. The old object is deleted when the new one lands, so there is no rollback and no copy of the previous cut on our side. Keep the previous export on your own drives before you swap. If losing it would be a problem, upload the new cut as its own film instead.
They use the same code they already have. Visibility and the access code are fields on the record that a replace never touches, so a password-gated link stays gated, an unlisted link stays unlisted and an embed-only film stays embed only. If you want the old audience locked out, that is a separate change to the access itself.
Not for a same-size swap. The quota check on a replace reuses the size of the file being replaced, so a v2 that weighs about what v1 weighed does not need headroom for both. A much heavier export still has to fit the plan, and the ceiling is 50 GB for any single file.
No. Nothing is transcoded here and there is no rendition ladder, so the film is live the moment the upload finishes and the first frame a client sees is the file you sent. There is no window in which a viewer gets a lower-quality version of the new cut.
The review room holds the versions, the timecoded comments and the approvals. When the cut is final, the same link carries the new file to everyone who already has it.