What Streaming Compression Does to Cinematography
Dune's desert gradients, The Batman's shadows, every foggy scene ever shot — compression quietly rewrites the look of films. What's actually happening to the image.
A cinematographer spends months lighting a shadow. A compression algorithm spends forty milliseconds deciding most of it is surplus. What streams out the other end is the film you meant to watch — minus the parts the encoder couldn’t afford. Understanding what gets cut explains a lot about why some films look transcendent on disc and vaguely muddy on streaming.
The Bitrate Budget Problem
Streaming is an economy. Every platform allocates a bitrate budget per title, and encoders spend it where the algorithm thinks detail matters. The casualties are predictable:
Gradients. Smooth transitions — skies, fog, the soft falloff across a face — are the first thing quantizers sacrifice. The result is banding: instead of a seamless dusk, you get visible terraces of color. It’s why shot-in-fog films suffer most on streaming. The desert gradients in Dune were a famously hard encode; early streams visibly banded where the Blu-ray didn’t.
Grain. Film grain — or digital noise added to mimic it — is random detail that changes every frame. To an encoder it looks identical to sensor noise: expensive, information-free, disposable. So grain gets smeared away or, worse, blocks up into a crawling texture that was never in the negative. Directors who shoot on film partly for that texture get a different film than they approved.
Shadow detail. Encoders allocate fewer bits to dark regions because the eye is less sensitive there — a sensible heuristic until a cinematographer builds an entire visual language out of near-blackness. The Batman’s shadow-heavy look was divisive partly because streams crushed its deliberately crushed blacks into mush.
Why It’s Worse in Motion
Encoders spend bits on change. Static, textured frames hold detail; the moment the camera moves, the budget gets spread across the whole frame and fine detail dissolves. A whip-pan across a detailed landscape is the toughest thing to encode; it’s why action sequences “look worse” than the dialogue scenes around them, even at the same nominal quality tier.
The Codec Hierarchy Helps — Partly
Newer codecs buy back some of the loss:
- AV1 handles film grain explicitly — AV1 encoders can strip grain at encode time and re-synthesize it at decode, preserving the look without paying the randomness tax
- HEVC is meaningfully better than H.264 at gradients and dark regions
- But a starving bitrate budget defeats any codec; a 4 Mbps AV1 stream still can’t preserve what needs 12
What You Can Actually Do
- Quality tier up when it matters. If a film’s look is the point — anything by Deakins, Lubezki, or shot on film — force the highest tier before the film starts; don’t let adaptive bitrate decide mid-scene.
- Wired > Wi-Fi for film nights. A sustained 25 Mbps pipe keeps the encoder in its generous mode instead of punting you down the bitrate ladder.
- Physical media still exists. For the films you love most, a 4K disc at ~100 Mbps is still the reference. We’ve logged the streaming-vs-disc comparisons that surprised us most at NungX technology notes.
Compression isn’t evil — it’s what makes streaming possible at all. But it’s worth knowing that the difference between “the film looked flat” and “the film looked like the cinema” is often just how many megabits the platform was willing to spend on you.