Why Two 1080p Videos Compress So Differently

Key Takeaways

• Resolution describes the pixel grid. It does not tell you how much detail survives compression.
• At the same bitrate, clean, mostly static footage is usually easier to preserve than noisy footage with complex motion.
• Constant-quality encoding allows file sizes to vary; a bitrate target allows quality to vary with scene complexity.
• GOP length changes prediction opportunities and seeking behavior, but a longer GOP does not automatically cause blur.

Two clips can both be 1920 × 1080, use the same encoder settings, and look very different after export. A slide presentation may remain crisp while a handheld night scene becomes soft and blocky. The encoder faces different problems even though the output dimensions match.

The useful question is: how much new, unpredictable information must each frame preserve?

Same pixel count, different work

Imagine a locked camera filming a plain wall with one person talking. Much of the picture stays similar between frames. Now imagine water splashing through leaves while the camera shakes. Both clips are 1080p, but the second contains many more changing details.

Motion prediction can reuse information, but it still needs to describe what prediction gets wrong. Occlusions, changing reflections and fine moving textures make that harder. Random sensor noise adds changes that do not follow the scene’s motion. These factors are described in MDN’s guide to video encoding complexity.

A static screen recording is often easy to compress, yet small text remains sensitive to damage. Scrolling a dense spreadsheet can be much harder than holding one slide on screen. “Screen recording” and “camera footage” are not universal difficulty rankings.

A bitrate budget is shared across frames

A rough comparison is bits per pixel per frame:

video bitrate ÷ (width × height × frames per second)

For an illustrative 6,000,000-bit/s video stream at 1920 × 1080 and 30 fps, the result is about 0.096. At 60 fps with the same bitrate, it is about 0.048. These are arithmetic examples, not quality recommendations or measured results.

This average is not a literal allowance for each pixel. Encoders distribute bits unevenly across time and regions. Adjacent frames at a higher frame rate can also be more similar, so doubling frame rate does not necessarily require twice the bitrate for comparable perceived quality.

Use the number to notice differences in the budget, not to predict quality across codecs, presets or kinds of footage. A noisy clip and a clean clip can have identical values and still look very different.

GOP structure: how frames borrow information

A group of pictures, or GOP, organizes coded pictures around intra-coded pictures and predictive pictures. In a simplified description:

• I pictures use prediction within the picture rather than referencing another picture.
• P pictures can predict from earlier reference pictures in decoding order.
• B pictures can use two reference lists, often drawing on pictures before and after them in display order.

These are frame-level shorthand; predictive pictures can also contain intra-coded blocks. H.264’s coding rules are defined by the ITU-T H.264 specification.

Longer GOPs can reduce the frequency of expensive intra pictures. Shorter intervals can help random access and recovery, while consuming more of a constrained bit budget. An I picture is not always an independent restart point; H.264 IDR pictures provide a stronger reset of reference dependencies.

Fast motion does not automatically make a long GOP smear. Lookahead, reference selection, scene-cut decisions and available bitrate matter. At a tight bitrate, complex action may lose detail regardless of GOP length. Forcing frequent keyframes can even leave fewer bits for intervening pictures. Keep the encoder’s defaults initially unless a playback or editing requirement calls for a particular interval.

Noise and grain compete with useful detail

A dim recording may contain fine, changing noise across almost the entire frame. Preserving that variation costs bits. Under a limited budget, the result may lose texture or show blocks; at a quality-oriented setting, the encoder may instead produce a larger file.

Light denoising can help, but compare faces, hair and fabric before accepting it. Aggressive filtering can erase genuine detail and leave motion trails. Intentional film grain may be part of the image you want to retain. Choose whether to preserve it before trying to make two exports the same size.

Why the same CRF produces different file sizes

With libx264, CRF is a quality-oriented rate-control mode: lower values generally retain more detail and require more data. It does not guarantee equal perceived quality across every scene. A bitrate target instead constrains the data budget, so harder scenes may show more loss. FFmpeg documents CRF, presets and GOP controls in its libx264 encoder options.

For an input with an audio track, a starting example is:

ffmpeg -i input.mp4 -c:v libx264 -preset medium -crf 20 -c:a aac -b:a 128k output.mp4

This requires an FFmpeg build with libx264. It re-encodes video and audio without explicitly resizing the picture or setting a new frame rate. The CRF value is a starting point to inspect, not a promised outcome. CRF numbers are not interchangeable across different encoders.

Diagnose a blurry export with a short comparison

1. Check the source first. Compare the same timestamp at the same display size. Existing blur and compression damage will not disappear through another lossy export.
2. Choose a difficult sample. Include movement, dark areas, fine textures or small text that matters to you.
3. Change one setting. For libx264 CRF, try a lower value; for a bitrate-controlled export, allow more bitrate. Keep the other settings stable.
4. Inspect playback and paused frames. Watch for texture disappearing during movement as well as damage to individual frames.
5. Match the setting to the goal. Let file size vary when visual preservation matters most. If storage or delivery imposes a firm budget, accept that difficult footage may need a different compromise.

Two 1080p exports should be judged by what they preserve from their own sources. Equal dimensions and equal settings alone cannot make their compression challenges equal.