Why Are My Video Clips Blurry? Export Fixes Explained

Four things called blurry

“Blurry” is not one problem. It is at least four, they look almost identical on a phone screen, and they have nothing in common underneath – which is why the usual advice of raising the export bitrate works for some people and does nothing at all for everyone else.

The middle row is the whole diagnosis, and checking it takes about ten seconds. Pause the clip on a still frame. If it is soft while nothing is moving, you are looking at missing pixels and no export setting will help. If it is crisp paused and falls apart in motion, that is compression, and compression is the one case where the export dialog genuinely matters. Blocky edges and banded gradients point at a file that has been encoded more than once. And if it is bad for a second and then fixes itself, nothing is wrong with your clip at all.

Try it: the pixels you never had

The most common cause of a blurry clip is the least discussed, because it does not feel like a cause: the footage never contained enough detail for the size you are posting at. This is arithmetic rather than opinion, so it is worth doing rather than reading.

Two results stand out. A 1080p recording cropped to vertical gives you 608 pixels across, which is being stretched by nearly two before anyone sees it – and 1080p is what most people record. The same crop from a 4K source lands comfortably the other side of the line, so the final step is a downscale, which looks sharper than no scaling at all because it averages real detail together rather than inventing it.

This is also the honest answer to the most common complaint about clipping, which is that vertical clips look worse than the video they came from. They do, and the reason is not the tool. Converting horizontal video to vertical covers what the crop costs and the ways around it, including the ones that keep the clip wide.

Why movement falls apart

If your clip is sharp when paused and turns to porridge when the camera moves, you have found compression doing what compression does.

A video codec does not store every frame. It stores what changed, which is why a talking head against a plain wall compresses beautifully and a handheld shot of a crowd does not. When the picture changes faster than the bitrate can describe it, the encoder degrades detail to stay inside its budget, and it does that exactly where your eye is already busy.

So the fixes for motion softness are mostly not settings. Reduce what changes. Stabilise the shot or lock the camera off. Add light, because noise is random change in every pixel and the encoder cannot tell it apart from detail. Cut on the motion rather than sitting through it. This is why gaming clips and stream highlights suffer worst – fast camera movement, particle effects and busy interfaces are close to a worst case – and why podcast clips almost never look soft.

Raising the bitrate does help here, genuinely, unlike in the upscaling case. It just helps less than people expect, for the reason in the next section.

Your file is not what people watch

Here is the thing that reframes every export decision: you never deliver your video to a viewer. You deliver it to a transcoder. The platform takes your upload and builds its own set of versions at its own bitrates, and one of those is what plays in the feed.

Which means the goal of an export is not to be as large as possible. It is to be easy to compress: clean, stable, correctly sized, at the frame rate it was shot. A noisy, shaky 200 Mbps file and a clean 20 Mbps file arrive at the same transcoder, and the clean one comes out the other side looking better.

YouTube publishes what it wants to receive, which is a useful anchor even though it says nothing about what gets sent out. For SDR uploads it recommends an MP4 with the moov atom at the front, H.264 High Profile with a closed GOP and variable bitrate, AAC-LC audio at 48 kHz, the BT.709 colour space, and bitrates of roughly 8 Mbps at 1080p and 35–45 Mbps at 4K for standard frame rates – rising to 12 and 53–68 Mbps at high frame rates.

Upload recommendations read from YouTube’s recommended upload encoding settings on 20 September 2026. These describe what YouTube asks to receive, not what it serves. Claims about how a transcoder rewards particular uploads are folklore unless the platform states them, and other platforms document this far less clearly than YouTube does – check your own account.

What the dialog can honestly fix

With all of that in place, the export settings that actually matter are few and dull. They will not rescue footage that lacks the pixels, and they will stop you making a clip worse than it needs to be, which is most of what an export dialog is for.

• Export at the delivery size, onceDo not upscale in the dialog hoping for detail, and do not export small and let the platform enlarge it. One scale, to the size you are actually posting.
• Keep the source frame rateYouTube says plainly that content “should be encoded and uploaded in the same frame rate it was recorded”. Changing it in the export is a common and avoidable cause of judder. (YouTube Help)
• H.264 High Profile in an MP4YouTube’s recommended upload settings are an MP4 container with the moov atom at the front, H.264 High Profile, progressive scan, closed GOP and variable bitrate. (YouTube Help)
• Use the recommended bitrate as a floorYouTube lists 8 Mbps for 1080p SDR at standard frame rates, 12 Mbps at high frame rates, and 35–45 Mbps for 4K. Going somewhat above is harmless; going far above mostly produces a larger file.
• Audio at 48 kHz, AACListed as AAC-LC at a 48 kHz sample rate. It has nothing to do with sharpness, and it is the setting people get wrong while worrying about video.
• Sharpen last, and barelyA light sharpening pass can help perceived detail. A heavy one gives the encoder high-contrast edges to describe, which costs bits and looks worse precisely where the picture moves.

Notice what is missing: any setting that adds detail. There is no such setting. Everything above either preserves what you have or avoids a second, unnecessary degradation, and that is the entire honest scope of an export dialog.

Stop re-encoding the same footage

Generation loss is the quiet one, because each individual step looks fine and the damage only becomes obvious three or four steps in.

Every time a video is encoded, the encoder discards what it judges you will not miss. Decode that result and encode it again and the second pass makes the same judgement about an image that has already been through it, this time including the artifacts of the first pass. Do it enough times and you get the familiar look: blocky edges, smeared movement, bands across what should be a smooth sky.

In practice the chain gets long without anyone deciding it should. A video is downloaded rather than exported from the source. It is screen-recorded because the download was awkward. It is trimmed and exported, then that export is reopened to add captions and exported again, then uploaded and re-encoded by the platform. That is five encodes, and only the last one was unavoidable.

The rule is simply one encode, from the original, as late as possible. Do the cutting, the crop, the captions and the titles in a single pass and export once at the end. It is also faster, which is the version of this argument that tends to stick.

Fixing it where it broke

Almost everything on this page points backwards, to a decision made before the clip existed. These are the habits that stop the problem happening, in rough order of how much difference they make:

• Record 4K if you will cropThe single highest-value habit on this page. A vertical crop of 4K leaves 1,215 pixels across; the same crop of 1080p leaves 608. Nothing downstream recovers that difference.
• Keep the original fileEdit from the camera file or the original download, never from a copy you already exported. Each additional encode costs something and none of it comes back.
• Do not screen-record a videoRecording a playing video is an encode of an encode, usually at a lower resolution and a mismatched frame rate. Get the real file if there is any way to.
• More light, less noiseNoise looks like detail to an encoder, so it spends bits describing grain instead of faces. A brighter scene compresses better and therefore looks sharper at the same bitrate.
• Stabilise, or stop movingCamera shake changes every pixel in the frame, which is the most expensive thing you can ask compression to encode. A locked-off shot is worth more than a higher bitrate.
• Match frame rates end to endRecord, edit and export at the same rate. Converting 30 to 25, or 60 to 30 badly, produces a judder people read as softness even when the frames are sharp.

The first one is worth more than the other five together. If you record at 4K and crop vertically, you are downscaling at the end and the whole subject of this page mostly goes away. If you record at 1080p and crop vertically, no export setting, no bitrate and no upscaler fully recovers it – you are enlarging 608 pixels and hoping, and the viewer can tell.

FAQ

Related: horizontal to vertical video, short video hooks and what video clipping is.