Nobody sets out to learn video compression. They set out to send a 43-second clip to a group chat, Discord refuses it, and half an hour later they know what a bitrate is. This post is the short version of that half hour.
What the limit actually is
Discord’s free upload ceiling has moved. It sat at 8 MB for years — long enough that “compress video to 8mb” is still how people phrase the problem in search engines — and it has since been raised. Nitro tiers raise it considerably further, and server boost levels have historically affected it too.
Because it has changed before and may change again, do not trust a number on a web page, including this one. Attempt the upload: Discord tells you the maximum it will accept from your account in that server, and that figure is authoritative in a way that a blog post cannot be.
Then compress to roughly 10% under it. Discord measures bytes; your file manager may be showing you megabytes computed a slightly different way; and being 40 KB over a hard limit fails exactly as completely as being 40 MB over. The headroom costs you nothing you can see.
Why gameplay clips are the worst case
Most video compresses well because most video barely changes from frame to frame. A video encoder’s central trick is describing a frame as a small difference from the one before it, and on a talking head or a screen recording that difference is almost nothing.
Gameplay defeats all of it at once. The camera swings constantly, so the entire frame moves. Foliage, particles, smoke and crowds fill the picture with fine detail that is expensive to describe and impossible to predict. A HUD sits on top with sharp-edged text and numbers that change every frame. And capture software — OBS, ShadowPlay, the Xbox Game Bar — records at a generous bitrate by design, because a moment you cannot re-take is worth spending disk space on.
The result is that a minute of 1080p60 capture at 200 MB is entirely normal, and the usual advice of “just lower the bitrate” runs out of road much sooner than it does on other content.
Workaround 1: compress it properly
The single most effective thing you can do is drop the resolution to 720p, and it is the step people resist hardest. The reasoning feels wrong — you are deliberately throwing away detail — but the arithmetic is clear. A 1080p frame is 2.07 million pixels and a 720p frame is 0.92 million, so the same bitrate is spread across 44% as many pixels and each one gets more than twice the attention.
More to the point, the clip is going to play in a Discord message on someone’s laptop, a few hundred pixels wide, and often on a phone. The 1080p detail was never reaching their eyes even when it reached their client.
After that, set a target size rather than fiddling with a quality slider. Target-size mode takes your ceiling in megabytes, divides by the clip length, subtracts the audio and encodes to what is left — then measures the result and re-encodes at a corrected bitrate if the encoder overshot. The Discord page opens with a 9 MB ceiling already set.
Workaround 2: trim it
This is the one that is almost always right and almost never tried.
Nearly every clip worth sharing has a fifteen-second version. The interesting thing happened at 0:38; the first thirty seconds are you walking toward it. Cutting those thirty seconds does not degrade a single frame of the part anyone cares about, and it doubles the bitrate available to what remains.
Trimming is the only genuinely lossless way to reduce a video file’s size. Everything else is a trade.
You do not need to open an editor first. Set the start and end points in the compressor, and target-size mode divides your ceiling by the length that survives the cut rather than the length you started with, which is the whole reason trimming and compressing in one pass beats doing them in two tools.
Workaround 3: link instead of attach
Upload the video somewhere else and paste the URL. Discord will embed a player for most of the common hosts, so the experience in the chat is close to identical, and the size limit stops existing.
The trade-off is that the clip now lives on a third-party service under someone else’s retention policy, and that the link can outlive the conversation or die before it. For a throwaway clip in a group chat, that is fine. For anything you would rather not have hosted publicly, it is not — and that is exactly the case where compressing locally is worth the effort.
Workaround 4: pay for Nitro
If you hit this limit weekly rather than annually, Nitro raises the ceiling and the problem goes away. That is a legitimate answer and it is worth naming, because a page full of workarounds that never mentions the paid option is being coy about something obvious.
When the clip genuinely will not fit
Below roughly 150 kbps, H.264 cannot hold a coherent picture together at any resolution. You get blocking, smearing, and frames that visibly rebuild themselves after every cut. Push past that point and you produce a small file nobody can watch, which helps no one.
For a 9 MB ceiling, that wall arrives at about six minutes with the audio still in, or eight if you strip it. Anything longer and the honest options are trim, link or Nitro — not “compress harder”.
One thing worth knowing about privacy
Gameplay clips and screen recordings are not neutral files. They routinely capture a desktop with open windows behind the game, a browser tab bar, a notification sliding in, a voice channel with real names in it, or a Discord username list down the side of the frame.
Uploading that to a free compression service means handing a stranger a recording of your desktop in exchange for a smaller file. The compressor on this site runs FFmpeg compiled to WebAssembly inside your own browser tab, so nothing is transmitted at all — which for this particular category of file is worth more than it sounds.
The short version
Check the real limit by attempting the upload. Drop to 720p. Set a target about 10% under the limit. If it still will not fit, trim the clip before you compress harder. And if the same thing happens every week, either pay for Nitro or get comfortable with links.