1080p encoding settings that hold up under review
1080p remains the workhorse delivery resolution. These are the parameter sets we publish as starting points, each paired with the reasoning that justifies it and the conditions under which it should be abandoned.
Full high definition at 1920 by 1080 is the resolution most catalogues still serve to most viewers most of the time. It tolerates moderately aggressive compression because typical viewing distance hides fine detail loss, but it punishes careless scaling and careless audio, both of which are audible or visible long before the video bitrate becomes the limiting factor.
Before choosing parameters, fix three inputs. The frame rate should match the source rather than a nominal broadcast value, because inventing frames costs bits and can introduce judder. The colour space should match the source and the display pipeline, since an unnecessary conversion round trip loses precision. And the audio should be handled at a bitrate appropriate to its content, because a low audio bitrate is noticed immediately even when the picture is flawless.
| Parameter | Value | Reason |
|---|---|---|
| quality mode | crf 21 | balanced rung for general delivery |
| preset | slow | recovers efficiency at acceptable speed |
| GOP length | 48 | two-second seek grid at 24 fps |
| B-frames | 3 | cheap deltas without deep reordering |
| profile | high | broad decoder support |
| level | 4.1 | covers 1080p at common frame rates |
| Parameter | Value | Reason |
|---|---|---|
| quality mode | crf 23 | equivalent percept to H.264 crf 21 |
| preset | medium | HEVC presets are costlier per step |
| GOP length | 48 | consistent seek grid across codecs |
| profile | main | 8-bit 4:2:0 compatibility |
| level | 4.1 | matches the H.264 tier above |
| Parameter | Value | Reason |
|---|---|---|
| quality mode | crf 30 | AV1 scale differs from x264 and x265 |
| speed | 6 | throughput acceptable for batch runs |
| GOP length | 48 | aligned across the codec mix |
| tile columns | 2 | parallel decode on multi-core clients |
Why these numbers and not others
The quality values follow the reasoning set out in the CRF and RF module: pick the lowest value that survives both an objective measurement and an eyes-on check at the intended viewing distance. The preset trades time for efficiency, and on 1080p the step from a fast preset to a slow preset is usually worth taking for masters while a medium preset is the pragmatic choice for volume.
Audio at 1080p
Stereo programme audio is comfortable in the range of 128 to 192 kilobits per second with a modern perceptual codec, and dialogue-heavy material tolerates the lower end. Multichannel content needs proportionally more, and loudness normalisation should be applied once, deliberately, during mastering rather than inherited accidentally from the source.
Failure modes to watch
Two mistakes recur. The first is upscaling to 1080p from a smaller source, which spends bits describing interpolated detail and leaves the impression of softness without any compression cause. The second is frame-rate conversion by duplication, which inflates the bitrate and creates unnatural motion. Both are corrected upstream in the filter graph and cannot be repaired later.
Finally, test the whole chain on a target device. A parameter set that measures well on a desktop player can still fail on a physical television if any of the declared profile, level or colour metadata falls outside what the display advertises. The measurement is necessary but not sufficient; the device test is the final word.