How AVIF Supports 10,000 Nits HDR Luminance
The AV1 Image File Format (AVIF) enables true High Dynamic Range (HDR) content delivery by natively accommodating mastering luminance targets up to 10,000 nits (candelas per square meter). It achieves this capability by leveraging the modern video coding standards of AV1, specifically through high bit-depth architectures, advanced non-linear transfer functions, and standardized HDR metadata signaling. This framework ensures that images captured and mastered for ultra-bright displays retain their dynamic range and color volume without clipping or degradation.
The Perceptual Quantizer (SMPTE ST 2084)
The primary mechanism that allows AVIF to scale up to 10,000 nits is its support for the Perceptual Quantizer (PQ) Electro-Optical Transfer Function (EOTF), standardized as SMPTE ST 2084. Traditional gamma curves (such as sRGB or Rec. 709) are relative curves anchored to an assumed maximum display luminance of roughly 80 to 100 nits.
In contrast, PQ is an absolute transfer function designed specifically to map digital code values directly to absolute luminance levels spanning from 0.0001 nits up to a theoretical maximum of 10,000 nits. When an image is encoded using the PQ curve, each integer value directly corresponds to a precise physical brightness level, allowing AVIF to define peak highlights anywhere within the 10,000-nit envelope.
10-Bit and 12-Bit Encoding Profiles
Covering the vast luminance range between 0.0001 and 10,000 nits requires fine gradation to prevent visual artifacts like contouring and color banding. AVIF supports AV1’s Main and Professional profiles, which offer:
- 10-bit color depth: Provides 1,024 discrete steps per color channel.
- 12-bit color depth: Provides 4,096 discrete steps per color channel.
When paired with the perceptual efficiency of the ST 2084 curve, a 10-bit or 12-bit sample depth distributes code values in alignment with human visual contrast sensitivity. This ensures that the wide dynamic range stretching to 10,000 nits is smoothly rendered without visible quantization errors.
Signaling via CICP (Color Information Code Points)
AVIF communicates color and luminance properties through
Coding-Independent Code Points (CICP), defined in ITU-T H.273 and stored
in the container's colr (Color Information) box. To achieve
the full range of modern HDR, AVIF can set:
- Color Primaries: Set to ITU-R BT.2020 for an ultra-wide color gamut.
- Transfer Characteristics: Set to Value 16 (SMPTE ST 2084 / PQ) to trigger 10,000-nit luminance decoding.
- Matrix Coefficients: Standardized color space transformation (e.g., BT.2020 non-constant luminance).
By parsing these values, an image decoder immediately recognizes that the pixel values correspond to an absolute luminance scale targeting the PQ curve rather than an arbitrary SDR curve.
Mastering Display Color Volume (MDCV) Metadata
While the PQ curve provides the mathematical capability to reach 10,000 nits, the actual mastering environment parameters are carried via Mastering Display Color Volume (MDCV) metadata, standardized under SMPTE ST 2086.
Within the AVIF format (derived from the ISOBMFF / HEIF structure),
this metadata is stored in the mdcv item property box or
within the AV1 sequence header Open Bitstream Units (OBUs). It
explicitly defines:
- Maximum Display Mastering Luminance: Up to 10,000 nits (stored in units of 0.0001 cd/m²).
- Minimum Display Mastering Luminance: Down to 0.0001 nits.
- Mastering Primaries and White Point: The exact chromaticity coordinates of the monitor used by the colorist.
This metadata tells the rendering pipeline that the master was authored on a reference monitor capable of producing up to 10,000 nits, providing essential context for down-mapping.
Content Light Level Information (CLLI)
To assist target displays that cannot achieve 10,000 nits—which
includes most consumer panels—AVIF supports Content Light Level
Information (CLLI) metadata via the clli property box. It
contains two critical values:
- MaxCLL (Maximum Content Light Level): The absolute brightest single pixel present anywhere in the image, up to 10,000 nits.
- MaxFALL (Maximum Frame Average Light Level): The average brightness of the entire image area.
Consuming devices use MaxCLL and MaxFALL to execute tone mapping. If an AVIF image contains a 10,000-nit specular highlight, a display capable of only 1,000 nits uses this metadata to roll off the highlights smoothly and compress the dynamic range without severe clipping or loss of midtone contrast.