Client delivery
Deliver multiple aspect ratios to a client:
fifteen files, one approval per shape.
A modern commercial ships one edit in four or five shapes, and each shape in three lengths. The reframing is the easy half. The hard half is that nobody can tell the files apart once they are sitting on someone's desktop, and a client who approved the widescreen has never seen the vertical. This is the delivery system for that, not a reframing tutorial.
Video hosting for filmmakers · 5 GB free · Paid plans from USD 9/month
Updated September 2026
The actual problem
Reframing is not the hard part;
it is telling fifteen files apart.
Every editor can pull a vertical out of a wide frame. What breaks a job is what happens after the renders finish. The files leave your machine, land in a download folder next to eleven other jobs, get forwarded to a media buyer who was not on the review, and one of them gets scheduled. Nobody in that chain can tell v02 of the square from v03, and nobody knows that only the widescreen was ever actually approved.
15
Files in a routine package
Five shapes times three cut lengths, before anyone asks for a captions-burned-in pass.
1
Approval per shape
A client who signed off the widescreen has not looked at the vertical.
0
Re-encodes on the way through
The 9:16 arrives exactly as you graded it, at the data rate you exported.
Treat this as a delivery system with four parts: a list of shapes decided before the edit, a naming convention that survives a desktop, a safe area that holds in every crop, and one approval per shape. Each part exists because of a specific failure. The list stops the fifth shape arriving after the grade is locked. The names stop the wrong file going live. The safe area stops a lower third running off both edges of the vertical. The approvals stop a boom in frame being discovered by the client’s agency, in public, after the post is scheduled.
Before the edit, not after
A shape asked for after the lock
is a change order.
Ask for the media plan, not the ratios
Clients do not know their ratios; they know where the campaign runs. Ask which placements are booked and derive the shapes yourself. A plan with a site hero, a paid vertical and an organic feed post is three shapes. A plan that says social is not an answer, and quoting against it is how the fifth shape arrives free.
Write the list into the quote
Shapes times lengths is a number, and that number is your render queue, your review load and your storage. Put it in the quote as a count of deliverables. Anything added later is priced later. This is the only step that costs nothing and saves the most.
Protect for the crop at the shoot
The room to reframe is bought on the day, not found in post. Frame with the vertical box on the monitor and keep the boom, the stands and the crew out of the top and bottom of the wide frame, not just out of the wide composition.
Set the safe area in the timeline on day one
Titles, lower thirds and logos get built inside the area common to every shape you owe, before the first title is designed. Retrofitting a title pass across five shapes is a second graphics job.
The shapes
What each ratio is for,
and what a naive crop costs.
Pixel dimensions below are arithmetic from a single UHD finish, so every shape comes out of the same frame and the losses are directly comparable. YouTube states 16:9 as the standard aspect ratio for its player on a computer, with 3840 by 2160 as the 2160p rung of its recommended ladder, which is why that frame is the usual master.
Dimensions are arithmetic from a 3840 by 2160 finish, with the 2.39:1 height rounded to an even 1608 (3840 divided by 2.39 is 1606.7). Loss figures compare a straight centre crop of that frame against its full width: 2160 px of width for 1:1, 1728 for 4:5, and 2160 multiplied by 9/16, or 1215, for 9:16. Delivery dimensions are a separate decision from the crop: YouTube states that Shorts upload at a maximum resolution of 1080p and run up to three minutes, so a 2160 by 3840 vertical is more than that particular placement takes. Downscaling from the larger export is a render you control; upscaling a 1215 px crop is not.
Read the last column as a shot list problem, not an export problem. The 9:16 figure is the one that ends arguments: a centre crop of a UHD frame gives you 1215 pixels of width, so the vertical is either reframed shot by shot with real decisions about who stays in frame, or it is a soft upscale of a third of the picture. That is the case for finishing from a capture larger than the delivery, and the case for framing on the day with the vertical box switched on.
Names that survive a desktop
Client,
project, length, ratio, version.
A file name is the only piece of metadata that survives every step of the chain. It survives the download, the forward, the re-download, the drop into a scheduling tool and the person who was never in the review. Everything else, the folder structure, the link, the thread, gets lost at the first hop.
The convention
ACME_SpringLaunch_60s_16x9_v04.mp4
ACME_SpringLaunch_15s_9x16_v02.mp4
ACME_SpringLaunch_15s_9x16_v02_CAPTIONS.mp4
Client first, because the file lands next to other clients’ work. Project second. Cut length third, so every shape of the 15 sits together. Ratio fourth. Version last, and always last, because that is what changes most often.
Write the ratio with an x, not a colon
9x16, not 9:16. Microsoft lists the colon among the characters reserved by Windows file systems and not allowed in a file name, so a colon either gets rejected or gets silently rewritten somewhere in the chain. An x costs you nothing and travels everywhere.
Zero-pad the version
v02, not v2. Desktops sort alphabetically, so v10 lands between v1 and v2 the moment a job runs past nine rounds. Two digits fixes it for the life of the project.
No spaces, no accents, one separator
Underscores between fields, nothing else. Spaces become %20 the moment a file is linked, accents survive one system and get mangled by the next, and a mix of hyphens and underscores makes the name unsearchable. Pick one and hold it.
Never ship a file called final
final, final2 and FINAL_use_this are how the wrong crop goes live. The version number is the whole point of the convention. If a client needs to know which one is current, the approval on record answers it, not the file name.
Safe area
Build titles in the area
that survives every crop.
Take the shapes above as a set. The narrowest of them, the 9:16, keeps 1215 pixels of width out of 3840. The shortest, the 2.39:1, keeps 1608 pixels of height out of 2160. Overlay the two and the area common to every shape is a centred box of roughly 1215 by 1608, about a quarter of the widescreen frame. That box is where titles, lower thirds, logos and legal lines belong if you want one graphics pass instead of five.
It is a strict rule and it is worth the strictness. A lower third built to sit near the left edge of the widescreen is not clipped in the vertical, it is entirely absent. A legal line set in small type across the bottom of a 16:9 is unreadable the moment it is scaled into a 4:5. And burned-in captions need more room than you think, because full-screen vertical placements draw their own interface over the frame, so anything close to the bottom edge competes with it.
Two practical habits. Put the guide boxes in the timeline as a locked layer rather than in your head, so every title is checked as it is placed. And check the graphics pass in the narrowest shape first: if it reads in the 9:16, it reads everywhere, and the reverse is never true.
The review problem
A client who approved the widescreen
has not approved the vertical.
This is the failure that costs money, and it happens after everyone has relaxed. The widescreen is signed off, the package ships, and a week later the agency finds a boom at the top of the 9:16, or a product label cropped out of the square, on a post that is already live. The picture was never the same picture. It was only ever reviewed once.
One film per shape
Each shape is uploaded as its own film with its own link and its own approval on record. A different shape is a different edit, not a setting on the hero, so approving one says nothing about the others. That is the whole mechanism: the yes is attached to the picture it was given for.
Versions inside a shape
Revisions stack on the film they belong to. v03 of the vertical sits on v02 of the vertical, the earlier cut stays, and the notes written against it stay with it. A note about the square never resurfaces on the widescreen, because it was never on that film.
Timecoded notes
Feedback lands on the frame it was written against. On a multi-ratio package that is the difference between a usable note and a guess: the boom at 0:07 in the vertical is not the same moment or the same problem as anything in the wide.
The transport side matters less than people expect, with one exception. Nothing is re-encoded on the way in or on the way out, so the 9:16 the client watches is the file you rendered, at the data rate you rendered it at. That closes off the most common argument in a multi-ratio round, which is the client insisting the vertical looks softer than the hero. If it does, the export is the place to look, not the delivery. Uploads run to 5 TB per file, subject to your available storage, and they resume rather than restart, which matters more for the masters sitting alongside the package than for the package itself. The mechanics of rounds and sign-off are covered in the review and approval guide, and the render side of a placement package in the social cutdowns guide.
When you do not need any of this
If you deliver one shape to one placement, this system is overhead and it will cost you more than it saves. A solo editor cutting a single vertical for one feed needs a file and a link, and a naming convention with five fields is ceremony. The overhead starts paying for itself at three shapes, or at the point where more than one person has to say yes. Below that, send the file.
Questions
Frequently asked
How many files is a multi-ratio delivery, really?
Count it before you quote it. Five shapes times three cut lengths is fifteen renders, and that is before a captions-burned-in pass or a version with the end card stripped. Most jobs land between nine and twenty files. The number is why the naming convention and the approval trail matter: at fifteen files, the failure mode is not a bad crop, it is the right crop with the wrong version number going live.
Does approving the 16:9 approve the other ratios?
No, and this is the single most expensive assumption in multi-ratio delivery. The vertical is a different picture: different headroom, different edges, different things visible at the top of frame. A boom, a light stand or a second person half out of shot can be clean in the widescreen and obvious in the 9:16. Each shape needs its own look and its own yes on record, which is why each one is uploaded as its own film with its own approval rather than as a setting on the hero.
Can I just crop the 16:9 to 9:16?
Arithmetically you can, and that is the problem. A straight centre crop of a 3840 by 2160 frame to 9:16 is 1215 pixels wide, about 32 percent of the original width, so roughly 68 percent of the picture is discarded and what is left has to be scaled up to fill a 2160 by 3840 delivery. It works on a locked-off single. It falls apart on a two-shot, on any move, and on any frame where the subject is not dead centre. Reframe shot by shot, or protect for the crop at the shoot.
What should the file name look like?
Client, project, cut length, ratio, version, in that order, with no spaces: ACME_SpringLaunch_15s_9x16_v03.mp4. That order sorts correctly on any desktop, groups every variant of the same cut together, and survives being downloaded, emailed and re-downloaded. Write the ratio with an x rather than a colon, because the colon is a reserved character that Windows does not allow in a file name. Zero-pad the version so v10 does not sort ahead of v02, and never ship a file called final.
Should each ratio be a separate upload or a version of the same film?
A separate film. A different shape is a different edit, with its own reframing decisions, its own notes and its own approval. Versions are for revisions inside one shape: v02 of the vertical stacks on v01 of the vertical, which keeps the earlier cut and the notes written against it. Getting that split right is what stops a note about the square resurfacing on the widescreen.
Does the vertical lose quality between my timeline and the client?
Not here. The player streams the exact bytes uploaded, with no transcode on the way in and none on the way out, so a 9:16 export arrives exactly as graded, at the data rate you rendered it at. What happens after the client publishes is a different matter: every publishing platform re-encodes on upload, and no host or setting changes that. Say so before the campaign runs.
How large can the files be?
Uploads run to 5 TB per file, subject to the storage available on your plan, and they are resumable, so an interrupted transfer picks up where it stopped. A fifteen-file ratio package is nowhere near that ceiling; it matters for the masters you keep alongside it.
One film per shape, one approval per shape
Each shape is its own film with its own versions and its own approval on record. Timecoded notes land on the frame they were written against, and the file the client downloads is the file you exported, because nothing is re-encoded on the way in or on the way out.
Sources
- 1.YouTube Help: video specifications and recommended resolutions (16:9 stated as the standard aspect ratio on a computer, with the 16:9 resolution ladder from 426x240 to 7680x4320; fetched 9 September 2026)
- 2.YouTube Help: create Shorts (Shorts up to 3 minutes long, uploads at a maximum resolution of 1080p; fetched 9 September 2026)
- 3.Microsoft Learn: Naming Files, Paths, and Namespaces (reserved characters not allowed in a file name, the colon among them; fetched 9 September 2026)
- 4.uncompressed.io: the client review room (versions per film, per-version approvals and timecoded notes; verified in product code 9 September 2026)
- 5.uncompressed.io pricing: plans, storage and unmetered viewing (storage is the only meter; views, viewers and uploads are not metered; the 5 TB per-file upload ceiling, subject to available storage, read in app/dashboard/Uploader.tsx; verified 9 September 2026)