Downloads Were Fast—But My VPN Couldn’t Upload the Client’s Video

The client’s reference files downloaded in minutes. My finished video did not seem capable of leaving the laptop at all. The upload climbed to 3 percent, stopped, then estimated that a 1. (Openvpn) GB export would take nine hours. I had forty-five minutes before the campaign team in Berlin began its review. I blamed the cloud service, cancelled the transfer and restarted it in another browser. The estimate briefly fell to twenty minutes, then climbed back into hours.

I was working from a shared studio in Tehran, a few weeks after international internet access had begun returning following an 88-day shutdown. Iran’s president ordered the reopening on May 25, 2026, but the first days of reconnection remained uneven. Businesses were still dealing with the consequences of months without dependable access to the wider internet. (Reuters)

For a video editor, “the internet is back” is not a complete description.

I could open the client’s project board. I could download music, graphics and a 600 MB folder of reference clips. Messages appeared with only a short delay. Yet the moment I tried to send the finished export in the other direction, the connection collapsed.

That was the failure I needed to solve. I did not need a VPN that looked fast while receiving files. I needed one that could keep a large upload moving until the client had it.

I first blamed the cloud service

The upload page seemed like the obvious suspect.

I switched from the client’s cloud workspace to Google Drive. The file moved quickly for a few seconds, then slowed to less than 100 KB per second.

I tried a direct transfer link.

The same thing happened.

Next, I compressed the export more aggressively, reducing it from 1. (Openvpn) GB to just under 800 MB. The smaller file should have been easier to send, but the connection remained so slow that I would still miss the review.

That changed the diagnosis. The file was large, but size was not the reason the transfer rate kept collapsing. I had removed almost half the data and ended up watching the same stalled progress bar.

Meanwhile, downloads remained normal. A revised soundtrack from the client arrived in less than a minute.

The imbalance was almost insulting.

A speed test made it stranger. With the VPN connected, there was enough download capacity for high-resolution video, but the upload result struggled to reach 1 Mbps. Without the VPN, the upstream number improved—but the foreign services I needed became inaccessible or unreliable.

Iran’s restrictions had already forced freelancers and businesses into similar compromises. During the shutdown, limited international access was introduced for some approved companies while ordinary users remained largely cut off. Reuters reported estimated economic losses as high as $80 million per day. (Reuters)

My upload was a small part of that larger disruption, but the immediate consequence was personal: if the outbound route failed, the client would not see the work.

The established provider kept giving me better downloads

The VPN I had started with was a large, familiar service. It had a long public history, extensive support resources and more server locations than I could realistically test.

Its automatic option connected quickly. Browsing felt normal, and downloads were genuinely fast.

I selected a geographically closer server, expecting the shorter route to improve the upload.

The transfer began at 2 MB per second, fell to 300 KB and then hovered below 100.

I chose another server.

Again, the upload started well and then collapsed.

The provider’s application showed no error. The tunnel remained connected. From the app’s point of view, nothing had failed.

From mine, the video was still trapped on the laptop.

Other VPN users have described the same practical split: downloads remain usable while upload speed drops toward zero. (Reddit) That was enough to expose my mistaken assumption. A healthy download did not prove that the entire VPN route was healthy.

The established service was bringing data in. It was not reliably carrying my work out.

Uploading stressed the route differently

A download and an upload may share the same connection, but they do not place the same demand on it.

Downloading the reference clips mostly moved data toward my laptop. Sending the final video required a continuous stream of encrypted traffic in the opposite direction. When that outbound path became unstable, the large transfer slowed even though incoming files still arrived normally.

Packet size can also become part of the problem. A VPN adds overhead, and some routes handle those larger encrypted packets poorly. Small requests continue working, while a bulk upload starts, slows and appears to hang. OpenVPN includes packet-size controls for troubleshooting this kind of failure. (Openvpn)

I could not observe the network operator’s or VPN provider’s internal filtering rules. I could see the pattern: pages opened, downloads completed and sustained outbound traffic repeatedly lost speed.

I considered changing the tunnel’s packet-size settings manually. That would mean testing one value, reconnecting, restarting the upload and repeating the process until something improved.

With fewer than thirty minutes remaining, that was no longer useful troubleshooting. It was another way to miss the deadline.

I needed the file delivered, not a technically satisfying explanation for why it was stuck.

Splitting the file only spread the problem around

My next workaround was less technical.

I divided the export into four archive files, hoping smaller uploads would finish before the connection slowed.

The first part reached 100 percent.

The second stopped at 61 percent.

When I restarted it, the cloud service created a duplicate. I now had one completed piece, two incomplete copies and a client who would need to download and reconstruct the project manually.

Even if all four parts arrived, I would simply be transferring my connection problem to the person waiting for the video.

I deleted the fragments.

That was the point where I stopped trying to make the established route fail more gracefully. Its download performance was useful, but download speed was no longer the comparison that mattered.

The only meaningful test was whether the complete export could leave the laptop without constant supervision.

The smaller app made the upload boring

I opened OnlydogVPN and selected the preset intended for a restrictive connection.

There was no long server list to work through. I connected, returned to the client’s original upload page and selected the unbroken 1. (Openvpn) GB export.

For the first minute, I watched closely.

The progress bar passed 5 percent.

Then 10.

The speed moved slightly, but it did not collapse. I answered two client messages, checked the final captions and looked back to find the transfer beyond 40 percent.

For the first time that afternoon, I could leave the progress bar alone.

The file reached 100 percent with fourteen minutes remaining. The client opened the link, confirmed that the video played and added one comment about the final title card.

The task that had caused the search was finished.

The smaller app combines traffic obfuscation with an HTTP/3-based connection. The obfuscation makes the tunnel less recognisable as conventional VPN traffic, while HTTP/3 handles packet loss and difficult network paths more efficiently. (IETF)

The visible result was simpler: the outbound route kept moving.

The upload did not need a spectacular peak speed. It needed enough steady speed to reach the end.

The revision tested it again

The client’s requested change was small: replace the final title card and export a new version.

Rendering took six minutes. The replacement file was almost as large as the first.

While it uploaded, the studio’s Wi-Fi weakened and the laptop switched to the phone hotspot I had left available as a backup. The transfer paused, resumed and continued without forcing me to begin again.

That second result solved a smaller but important problem. The first upload showed that the service could carry a large outbound transfer. The revision showed that it could preserve the work when the network underneath it changed.

The new version reached the client before the review began.

I closed the cloud page and realised that I had not run another speed test during the final half hour.

There had been no reason to.

Why the normal download speed misled me

After the deadline, I repeated the comparison with smaller files.

The established provider still downloaded quickly. On some servers, it received files faster than the smaller service.

Its uploads remained inconsistent.

That made the weakness in my original comparison obvious. Most people test a VPN by opening websites, playing video or downloading a file. Those actions reveal whether the incoming path is usable. They may say very little about a designer sending a project, a developer pushing a build or a journalist uploading footage.

The smaller service has fewer locations than the major provider, a shorter public history and fewer independent reviews. Those limitations matter when someone needs a specific regional endpoint or places the greatest weight on years of outside scrutiny.

They did not decide whether my video reached Berlin.

The large provider gave me strong download numbers and repeated upload stalls. Compressing and splitting the file reduced the amount of data but added work without repairing the route. The smaller app kept the complete transfer moving until the client could press play.

When VPN download speed looks normal but upload speed collapses, the useful connection is not the one that brings files in fastest. It is the one that lets your finished work leave.