Why your video looks worse after you upload it

Your file is not stored and served back. It is decoded, re-encoded into the platform's own formats, and then served to viewers from whichever version fits their connection and device. The copy anyone watches is at least one generation removed from what you exported, and often two.

That re-encode is not tuned for your video. It runs at scale, on a schedule, against settings chosen to serve billions of streams cheaply. It has no idea that the fine texture in your background is the point of the shot, and it will spend bits on whatever its heuristics say matters.

So the framing that helps is this: you are not delivering a finished video to a viewer. You are delivering a SOURCE to an encoder. The question stops being how good your export looks and becomes how well it survives being crushed by something that will not consult you.

There is some relief in that reframing. If your uploads have been coming out softer than your exports, it is not a verdict on your production and it was never something better taste would have avoided. It is a well-understood engineering problem with a short list of fixes, and once those are set correctly in a pipeline they stay set for every video you publish afterwards.

Why the same settings give different results

Two videos exported identically can come out visibly different, and the reason is almost always content rather than settings. Compression works by discarding information a viewer is unlikely to miss, and how much it can discard depends entirely on what is in the frame.

A static talking head over a plain wall compresses beautifully, because almost nothing changes between frames. Confetti, rain, water, fire, dense foliage, film grain and fast camera moves are the opposite: every pixel changes every frame, there is no redundancy to exploit, and the encoder runs out of bitrate and starts visibly giving up. This is why the failure looks like blocky mush in exactly the busiest, most expensive-looking parts of your video.

This is the most repeated piece of advice in the category and it is half true, which is worse than being wrong. The claim is that uploading 4K makes your 1080p stream look better. The reality is more specific than that, and following it blindly costs you.

There are two real effects. Platforms have historically allocated more generous bitrate to higher resolution tiers, so a 4K upload can be assigned a better encode. And downscaling averages pixels together, which cleans up noise and can produce a sharper-looking result at the delivery resolution.

But short-form vertical video is watched on a phone, where the delivered stream is a fraction of what a desktop viewer of a landscape video gets, and where the difference is far less visible than the advice implies. Meanwhile the 4K path costs you longer generation, longer export, longer upload and more storage on every single video, forever.

The native numbers differ per platform and change, so check them against the current resolution, aspect ratio and duration limits per platform

The export dialog has a lot of controls and about five of them change the outcome. The rest are either irrelevant to a platform upload or are decided for you by the re-encode.

• Codec. H.264 in an MP4 container is accepted everywhere and is the safe default. Newer codecs compress better at the same quality, but the file is being thrown away and re-encoded anyway, so their advantage is upload speed rather than final quality. Compatibility is worth more than efficiency here.
• Bitrate. The setting that matters most. You are not optimising for a small file. You are handing the encoder the most information you reasonably can. Generous is correct; the platform will take it from there.
• Frame rate. Match your source and keep it constant. Variable frame rate is a common source of audio drift and stutter after re-encode, and it comes out of screen recorders and phones by default.
• Colour. Standard dynamic range in Rec.709 unless you have a specific reason otherwise. HDR handed to a platform that tone-maps it badly produces washed-out or oversaturated results that look worse than SDR would have.
• Audio. AAC, stereo, 48 kHz, at a healthy bitrate. Audio is a small share of the file and the first thing viewers notice when it is bad.

Two-pass and why it usually is not worth it

Two-pass encoding analyses the video first and distributes bits more intelligently on the second pass. For a file that will be watched as-is, it is a genuine improvement. For a file about to be re-encoded by a platform, it roughly doubles your export time to hand a slightly better source to something that is going to discard most of it.

At one video a week, use it if you like. At a publishing cadence, the export time is real and the benefit is mostly theoretical. Spend the same effort on bitrate headroom instead, which costs nothing but disk.

Settings are the smaller half. The larger half is that some creative choices are expensive to compress and some are cheap, and choosing cheap ones is free quality.

• Slow the camera moves down. A slow push in preserves temporal redundancy between frames. A fast whip pan destroys it and forces the encoder to rebuild the entire frame repeatedly. This is the single biggest content-side lever.
• Avoid added grain and noise. Film grain is random by definition, which makes it incompressible. Grain added in post survives the encode as blocky mud, so it costs bitrate and delivers the opposite of the intended texture.
• Keep text large and high contrast. Fine text is the first thing to smear. Subtitles that are readable after compression are larger and bolder than subtitles that look right in your editor.
• Watch busy backgrounds. Dense foliage, crowds, water and particle effects all eat bitrate that your subject needed. Simplifying the background buys sharpness on the thing you actually wanted sharp.
• Be careful with hard cuts on every beat. Every cut forces a full keyframe. Cutting on every beat of a fast track means the encoder is constantly starting over, which is why heavily cut edits often look softest.

Caption styling has its own set of rules, which we covered in how to style subtitles that stay readable

Before changing anything, work out which end the problem is at. Most of the time people spend a week adjusting export settings for a problem that was introduced three steps earlier.

• Look at your export first, full screen. Not the editor preview, which is often a lower-quality proxy. Open the actual exported file at full size. A surprising share of upload complaints are export problems the creator never inspected.
• Check the source clips at full size. If a generated clip was soft to begin with, no export setting recovers it. Generation resolution is a ceiling that nothing downstream raises.
• Wait before judging the upload. Platforms serve a low-quality transcode first and finish the higher-quality versions afterwards. Judging a video in the first minutes after upload is judging a version that will be replaced.
• Force the highest quality in the player. Players default to adaptive streaming and will serve a lower rung on a weak connection. Compare like for like by pinning the quality manually before deciding your upload was ruined.

That third point catches more people than any technical setting. The video you looked at ninety seconds after upload is not the video your audience will see an hour later, and a lot of export-setting folklore comes from comparisons made against a transcode that had not finished.

That last point is the main technical argument in how to repurpose one video across platforms without stacking re-encodes