4K and HDR encoding settings: parameters and the reasoning behind them
Ultra high definition multiplies pixel count by four and frequently multiplies bit depth and dynamic range as well. The parameter sets below assume a clean 4K source and a delivery target that can decode what you send.
At 3840 by 2160 the encoder has four times as many blocks to describe as at 1080p, and the artefacts that were hidden at lower resolution become easier to see because each block covers proportionally more of the screen. Compression efficiency therefore matters more, not less, and the choice of codec has a larger visible consequence than it does on a full high definition encode.
Two source properties must be established before any quality value is chosen. Bit depth determines how much precision the encoder has available before quantisation begins, and high dynamic range determines how much of the tonal range must be preserved without banding. Both are inherited from the source and must survive the filter graph intact.
| Parameter | Value | Reason |
|---|---|---|
| quality mode | crf 24 | preserves tonal range in HDR grades |
| bit depth | 10 | guards against banding in gradients |
| chroma | 4:2:0 | standard for consumer delivery |
| profile | main10 | 10-bit main-tier decoding |
| level | 5.1 | 4K at up to 60 fps |
| GOP length | 48 | consistent segment seek grid |
| Parameter | Value | Reason |
|---|---|---|
| quality mode | crf 32 | AV1 scale for high-detail content |
| bit depth | 10 | mandatory for HDR preservation |
| tile layout | 2x2 | parallel decode across cores |
| film grain | synthesis | models noise instead of coding it |
| level | 5.1 | aligned with the HEVC tier |
| Parameter | Value | Reason |
|---|---|---|
| quality mode | crf 20 | compensates for weaker efficiency |
| profile | high | widest decoder support |
| level | 5.1 | required envelope for 4K |
| bitrate cap | moderate | keeps fallback transport predictable |
HDR metadata must survive
High dynamic range delivery depends on metadata that describes the transfer function and the mastering display. Static metadata travels with the stream and is broadly supported; dynamic metadata can adapt tone mapping per scene and rewards capable displays. Whichever is used, the filter graph must pass it through unchanged. A conversion that silently drops the metadata produces a picture that decodes but looks wrong on every correctly configured display.
Bit depth, banding and noise
Ten-bit encoding costs little more than eight-bit in storage and decode complexity on modern hardware, and it materially reduces banding in smooth gradients such as skies and studio backdrops. That alone is a strong argument for ten-bit wherever the decoder supports it. The counter-argument is device coverage, and the honest resolution is a codec-mixed ladder that serves ten-bit to capable clients and a compatible fallback to the rest.
Performance expectations
A 4K encode takes longer than the equivalent 1080p encode by more than the pixel ratio, because search effort grows with both resolution and quality target. Hardware encoders recover much of that time, at the rate-quality cost described in the GPU acceleration module. Plan capacity from measured throughput on representative material, not from vendor throughput claims measured on synthetic content.
Verify the delivered file on the actual display that will show it. Profile, level, bit depth, colour metadata and HDR signalling all have to agree before a 4K asset plays as intended, and the only conclusive test is playback on target hardware.