8-bit vs 10-bit: Color Depth, Banding and Why Encoders Prefer 10
Enlarge a sunset frame and the orange-to-blue gradient hides ring-shaped boundaries, as if watercolor blending were cut into stairs. That is not a bitrate problem — it is color depth. The 8 versus 10 bit gap is decided exactly where gradients live.
What color depth is: steps per channel
Color depth is how many levels each color channel records: 8-bit gives 2⁸ = 256 levels per channel, about 16.7 million RGB combinations; 10-bit gives 1024 levels, over a billion. Denser steps mean smaller jumps between adjacent levels and smoother gradients.
The key framing: depth governs precision *inside* a color transition — the thousands of intermediate shades between red and blue are exactly where depth earns its keep.
Where banding comes from
A sky fading from orange to blue may span over a hundred levels — at 8-bit’s 256 total, each step is a visible 0.4% jump at normal viewing distance: banding, the contour-line artifact. At 10-bit, each step shrinks to 0.1%, below perception.
Banding’s favorite habitats: large gradients (skies, light falloff, bokeh), shadow transitions (8-bit shadows are inherently low-precision), and smooth gradients in motion (moving bands are even more visible).
Why encoders prefer 10-bit: the internal-precision dividend
A counterintuitive fact: even when the source is 8-bit and the output must play on 8-bit devices, encoding through 10-bit internal processing reduces loss. x264/x265’s 10-bit modes compute quantization, transforms and motion compensation at higher precision, with smaller rounding error — at equal bitrate, 10-bit-encoded output shows less banding and cleaner shadows.
That is why the HEVC ecosystem defaults to Main10: compression efficiency and quality improve together, at the cost of compute. When a compression tool offers a 10-bit HEVC option, gradient-rich content should take it.
The compatibility reality
10-bit video (HEVC Main10, VP9 Profile 2, AV1) decodes on recent phones, TVs and GPUs without trouble; older devices (pre-2016 phones and boxes) may reject it or software-decode into stutter; browser support for 10-bit HEVC/AV1 varies by platform. The gap versus 8-bit is real but closing fast.
Delivery to unknown audiences still defaults to 8-bit H.264; 10-bit suits known-capable environments — your own TV, platforms that explicitly accept HEVC 10-bit.
What content deserves 10-bit
Biggest wins: large gradients (landscapes, sunsets, lighting, fog), shadow-heavy footage, animation and composites (vast flat gradients). Near-zero wins: normally lit talking heads, interviews, indoor documentary — too few gradient regions for the eye to catch.
A practical decision tree: 10-bit source (modern cameras, phone HDR) → process and archive in 10-bit to preserve precision; 8-bit source of mostly speech → stay 8-bit end to end, and don’t pay compute for gradients that don’t exist.
Three sources of banding, three fixes
Banding in a frame is not always depth: low bitrate (quantization crushing gradients into steps), chroma subsampling (4:2:0 halves color precision — its own guide), and true depth limits. Symptoms resemble each other; remedies differ. Check the file’s depth and bitrate with an inspection tool first, then treat the actual cause.