My name is Christian Gomez. I run MarketingSTRAT, a one-person studio in Quebec City that shoots corporate and industrial video, and I also built and operate uncompressed.io. So let me be direct about what this page is: it is first-party writing about my own product, by the person who made it. It is not a review, it is not independent, and you should read it with that in mind. My bias is not hidden. I would like you to use the thing I built.
The technical half of this page is true regardless. What a platform does to a graded master on upload, why it does it, and how the pricing of video hosting is actually constructed are all facts about the industry, not opinions about my software. If you read this and go back to what you already use, you will still know more about your own delivery chain than you did before.
I wrote it because I spent years sending client links and watching a grade I had approved on a calibrated monitor arrive on the other end looking like a different film. Everyone in this business has had the call: the client says the image looks noisy, or banded, or flat, and you have no honest way to tell them whether that is your work or the pipe it travelled through. That is the problem this page is about.
Start with the file you actually finish on. Apple's published ProRes data rates put ProRes 422 HQ at 1920x1080, 29.97 fps at roughly 220 Mbps, and ProRes 422 at roughly 147 Mbps. At UHD those numbers scale by about four. That is the file your colourist worked in and the file you signed off.
Now look at what a general-audience platform asks you to hand it. YouTube's published recommended upload encoding settings ask for about 8 Mbps at 1080p SDR and 35 to 45 Mbps at 2160p SDR. From a ProRes 422 HQ master to an 8 Mbps 1080p upload is roughly a twenty-five to one reduction, and that happens on your own machine, before the platform has touched anything.
Then the platform re-encodes what you gave it. It builds a ladder of renditions at different resolutions and bitrates so that a phone on cellular and a smart TV on fibre both get something playable. Your viewer is not watching your export. Your viewer is watching an encode of an encode, chosen at runtime by an algorithm reacting to their connection.
What a grade loses in that chain is specific and predictable. Delivery is almost always 8-bit 4:2:0, so the chroma resolution you had in a 10-bit 4:2:2 master is thrown away: soft warm shifts on skin, subtle splits between foreground and background, anything you did with a qualifier gets coarser. Gradients band. Night skies, smoke, fog, a slow fall-off on a studio backdrop, a sunset: these are the shots where you see stepping that was not in the master. Fine grain, whether it came from the sensor or from a grain plate you added on purpose, is what rate control hates most, so it either gets smeared into mush or eats the bit budget that should have gone to detail. High-motion, high-detail frames block up: foliage, water, rain, crowds, handheld coverage of a factory floor. And the shadow detail you protected at the low end is the first thing a bitrate ceiling gives away.
None of this is a platform being careless. It is the deal you accepted when you uploaded. The encode is the product.
It is worth being fair about this, because the transcode is not arbitrary and it is not a scheme.
Storage is the cheap part of video hosting. It has been getting cheaper for two decades and it is a fixed cost per file. The expensive parts are egress, the bandwidth of actually shipping bytes to viewers, which is metered per gigabyte and recurs on every single play, and transcoding, which is a heavy one-time compute cost the platform amortises over all future views. A platform that expects a video to be watched a hundred thousand times will happily spend real money encoding it once, because that spend buys a permanent reduction in the per-view bandwidth bill.
The second reason is reach. A public platform has to play on a five-year-old Android phone on a hotel connection. A 220 Mbps stream is not a difficult delivery problem for that viewer, it is an impossible one. Adaptive ladders exist so that everybody gets something rather than nobody getting anything.
So a general-audience platform is right to transcode. The mismatch shows up when filmmakers use a general-audience platform for an audience of four people. Master delivery, colour sign-off, a director in another city, an agency lead approving a final grade: that is not a hundred thousand views on cellular. It is a handful of people on decent machines and decent connections, each of whom has a professional reason to see the real thing. Optimising that job for phone-scale delivery costs you the only thing that mattered.
Because egress is the real cost, plans get priced on whatever the vendor can use as a proxy for egress. That is why the pricing pages in this category look the way they do. Some meter total bandwidth per month. Some meter plays or impressions. Some meter seats and put the features you need behind the seat count. Some do not meter at all and instead reserve the right to throttle.
The practical effect for a filmmaker is that the cost of your account is driven by how many people watch, which is the one variable you neither predict nor control. You send one link to one client contact. That contact forwards it internally to a distribution list. Forty people open it, several of them twice, one of them scrubs the timeline for ten minutes on a 4K rendition. You did not do anything, and you are now in a different pricing tier or watching a progress bar that never fills.
The throttle version is worse than the bill, because it fails in front of the client. The link does not tell them you hit a cap. It just stops working, on the day of the review, and you look like the problem.
I priced Uncompressed.io on storage instead, and there are no bandwidth caps. Storage is the variable the customer can see, control and reason about. You know roughly how many terabytes of finished work you keep. You do not know how many times a client's marketing department will open a link.
Separate from fidelity, there is the question of what a review tool has to do to survive contact with a real client. After enough rounds of notes, the list gets short and specific.
On Uncompressed.io the review link is the login. There is no signup for the reviewer, no password to reset, no invitation to accept. They click the link and they are in the room. Notes are timecoded, drawing on frame is built in, versions stack with their history, and approvals are recorded per version.
One more thing that belongs in this list even though it is not functional: there is no player branding. When you send a client a link, the page carries your work and your name, not a hosting company's logo sitting in the corner of your frame.
Here is the hour nobody bills for. Notes live in one window, the timeline lives in another, and an editor spends an hour reading a list, scrubbing to each timecode by hand, and typing markers. It is transcription work, it is error-prone, and a single mistyped timecode means a note gets applied to the wrong shot and comes back again on the next round.
Notes from Uncompressed.io return into DaVinci Resolve and Premiere as markers on the timeline, at the frame the client was actually looking at. The note text rides along with the marker.
The reason to build it that way rather than exporting a tidy PDF is simple: a note is only useful where the work happens. A PDF of notes is a document about the edit. A marker is in the edit. When the note is on the timeline, the editor works down the timeline and the round closes, instead of alt-tabbing between two applications and hoping nothing was missed.
Storage for finished video is really two different problems that get billed as one, and most people overpay by treating them the same.
Hot storage is for work in flight. It needs instant playback, random seek, many reads by many people, and it needs to be there the second a client clicks. It costs what it costs because performance is what you are buying.
Cold storage is for the archive. The delivered job you must keep for years because the client will eventually ask for a re-cut or a new aspect ratio. It gets read almost never. Cold tiers are priced far lower per terabyte per month, and the trade is retrieval: getting a file back takes time, and sometimes costs money.
| Hot | Cold | |
|---|---|---|
| What it is for | Active projects, review, delivery | Finished work you must keep |
| Access | Immediate, streamable | Delayed retrieval |
| Relative cost per TB | High | Much lower |
| Right time to use it | While the job is alive | After the job is closed |
The two mistakes are symmetrical. Paying hot prices for a five-year archive is the expensive one, and it is extremely common, because nobody ever schedules the cleanup. Putting an active project in a cold tier is the painful one, because the client is on the phone and you are waiting on a restore.
A workable rule: hot while the job is alive, plus a defined tail after delivery, say ninety days, then cold. And one thing I will say against my own interest: a hosting account is not a backup. It is a review and distribution layer that happens to hold a copy. Keep your own masters on your own drives, on tape, or on a local array, and treat any cloud copy as convenience rather than insurance.
This question comes up in every conversation now, and it deserves a non-marketing answer.
Almost every platform that hosts user-uploaded video needs a licence to your file, because storing it, transmitting it, making thumbnails from it and displaying it are all acts that copyright covers. A licence grant in the terms is not by itself sinister. What matters is its scope, and you find that in the terms of service, not on the pricing page.
When you read a hosting agreement, these are the parts that actually decide the answer. Does the grant stop at operating and providing the service, or does it extend to language like improving our products and services, developing new technologies, or machine learning. Is the grant sublicensable, which is the clause that permits handing your material to a third party. Does the grant survive deletion of the file and termination of your account. Is there a separate AI or data section, and if there is, is it opt-in or opt-out by default.
For Uncompressed.io, the contract says customer footage is never used to train AI models. I want to be precise about what that claim is and is not. It is a contractual commitment, which is the thing you can hold a company to and the thing it can be liable for. It is not a proof, because no customer can audit a running system from the outside, and anyone who tells you otherwise is selling you a feeling. What you can do is read what a company is willing to write into a binding document, and prefer the ones willing to write it plainly.
The whole product follows from one decision: no re-encode. The file you upload is the file that is stored and the file that plays. There is no transcode step, no ladder, no second-generation encode standing between your grade and the person approving it. Video hosting with no compression applied after the fact is unusual, and it is the entire reason the thing exists.
What that means in practice:
It is at uncompressed.io. The free tier is enough to put one real graded master through it and see, on your own footage, whether the difference is one your clients would notice.
I would rather write this than have you find out later.
If the video is public marketing content aimed at a broad audience watching on phones, you want a general-purpose platform, and I would tell a client the same. Adaptive ladders, global delivery, discovery, embedded players tuned for every device, captions and analytics as standard, and often no cost at all. Serving that audience a full-bitrate master would be worse for them, not better. Their connection would lose, and the thing they see would be a spinner.
Uncompressed.io is for the other end: master delivery, client review and approval, colour and finishing sign-off, QC before a broadcast or festival submission, handing an agency the actual file rather than a picture of it. Small audiences, high stakes, people with a professional reason to see exactly what you made.
A lot of studios run both, and that is the correct arrangement rather than a contradiction. Compressed hosting for the public cut, uncompressed hosting for the master, the review and the archive. The mistake is only ever using one tool for both jobs and then wondering why the grade does not survive the trip.
No transcode. The file you upload is stored as-is and played as-is, and the reviewer can download the original. There is no rendition ladder and no second-generation encode between your export and the person watching it.
On a decent connection and a reasonably modern machine, yes, that is the normal case. The honest limit is bandwidth: streaming a master at full bitrate means the connection has to keep up, so a weak link buffers rather than quietly dropping to a worse-looking version. That trade is deliberate. If a client is on hotel wifi and only needs to check pacing, send them a small file for that pass and use the master for the grade approval.
For master delivery, client review and approval, yes, and there is one-click Vimeo import so moving an existing library is quick. For public marketing video aimed at a large phone audience, a general platform with adaptive delivery is still the better tool. Plenty of studios use both.
No. The review link is the login. A client clicks the link and is in the review, with timecoded notes, drawing on frame, version history and approvals. No signup, no password, nothing to forward to IT.
No. It is 5 GB with no credit card and no time limit. It is enough to put one real graded master through the system and judge it on your own footage.
Paid plans start at 9 dollars per month for 100 GB and go up to 250 dollars per month for the top tier. Pricing is on storage, not bandwidth, so there are no bandwidth caps and the bill does not move because a client shared the link internally.
No. The contract states that customer footage is never used to train AI models. That is a binding commitment rather than a proof, and I would give any vendor the same test: read the licence grant in the terms of service, check whether it extends past operating the service, whether it is sublicensable, and whether it survives deletion of your files.
If any of this matches a problem you actually have, the free tier is the only sensible way to judge it: 5 GB, no credit card, no expiry, and no argument to have with me about whether the difference is visible. Put one graded master through it, send the link to the person whose approval you need, and see what comes back. It is at uncompressed.io. If it is not the right tool for the work you do, the section above on when a compressing platform is the better choice was written in good faith, and I would rather you use the right thing than my thing.
Book a call