Learn / FundamentalsWhat is bit depth in an image?
8-bit, 10-bit, 12-bit — the number that controls how finely a color channel can be divided, and the real explanation for why smooth gradients sometimes show visible steps.
"8-bit" and "10-bit" get thrown around like specs on a shopping checklist, but the underlying idea is simple: bit depth is how many distinct values a single color channel can hold. Everything else — banding, HDR headroom, why some formats need more bits than others — follows directly from that one number.
The actual math
A channel stored with n bits can represent 2n distinct values. 8-bit gives you 256 levels per channel (0–255) — the standard for JPEG, most PNGs, and most displays you've ever used. 10-bit gives you 1,024 levels per channel; 12-bit gives you 4,096. Multiply that across three channels (red, green, blue) and the jump is dramatic: 8-bit's ~16.7 million total colors versus 10-bit's ~1.07 billion.
Where you actually see the difference: banding
A smooth gradient — a sunset sky, a soft studio backdrop — is a continuous change in brightness. Stored at 8-bit, that continuous change gets rounded into 256 discrete steps per channel. Most of the time that's more than enough for the eye to perceive as smooth. But stretch a subtle gradient across a wide enough range, or push it through enough rounds of editing, and those discrete steps can become visible as faint bands — most noticeable in dark tones and smooth skies, where the eye is most sensitive to small brightness differences.
This is a real, verifiable thing about a specific file, not a guess — our
Inspect tool reads a JPEG's actual SOF marker (or the equivalent field in PNG/AVIF) to report the file's real bit depth, rather than assuming based on format or file extension.
Why HDR content wants more bits
HDR needs to represent both the normal SDR brightness range and real extra range above it — highlights that go well past what SDR white used to mean. Cramming a wider brightness range into the same 256 levels means each step covers a bigger perceptual jump, making banding far more likely. That's the practical reason PQ and HLG (HDR's two main transfer functions) are specified for 10-bit or higher, not 8-bit — see What is PQ? and What is HLG? for how each handles that range.
Gain-map HDR JPEGs (the format Headroom's
Encode tool produces) are a bit of an exception worth calling out honestly: both the primary and gain-map images are standard 8-bit JPEGs. The extra dynamic range doesn't come from a wider bit depth on either individual image — it comes from combining two 8-bit images mathematically at render time. It's a genuinely different mechanism from a native 10-bit HDR pixel format, not a smaller version of the same idea.
Checking a real file instead of guessing
Bit depth is exactly the kind of thing worth reading directly off a file rather than assuming from its extension or source. Inspect parses the real bytes — JPEG's SOF precision field, PNG's IHDR chunk, AVIF's pixi box — and reports GOOD when it can confirm the value directly, rather than guessing.
Questions people actually ask
What does '8-bit' actually mean?
It means each color channel (red, green, blue) is stored using 8 binary bits per pixel, giving 2^8 = 256 possible values per channel — 0 to 255. Combine three 8-bit channels and you get roughly 16.7 million possible colors per pixel.
Why do gradients sometimes show visible bands instead of a smooth blend?
Banding happens when a gradient's brightness range gets stretched across too few discrete steps for the eye to perceive it as continuous — most often 8-bit content that's been broadened (e.g. a subtle sky gradient) beyond what 256 levels can represent smoothly, especially in dark tones where the eye is most sensitive to small differences.
Do I need 10-bit for a normal photo?
Usually not for final delivery — 8-bit is plenty for most photographic content viewed on a standard display, which is why JPEG (8-bit) remains completely standard. 10-bit and higher matter more during editing (to avoid compounding rounding errors across many edits) and for HDR content, which needs more headroom to represent both the existing SDR range and real extra brightness above it.
Does a higher bit depth make a file bigger?
Yes, roughly proportionally — 10-bit per channel stores more data per pixel than 8-bit. How much that affects the final file size depends on the format and its compression; some formats (like PNG) scale fairly directly, others less so.
