Colour Science & Management
Modern colour grading has shifted from LUT-based workflows to colour-managed pipelines built on precise mathematical transforms. Understanding colour spaces, working spaces, and CSTs is now as fundamental as understanding the camera formats they connect — from sensor to screen.
01 — How colour is captured
Film and digital cameras both record light — but the way they do it is fundamentally different, and that difference is the reason the entire colour science pipeline exists.
Film: logarithmic by nature
Film captures light through a photochemical process. Silver halide crystals in the emulsion respond to photons: the more light, the more crystals darken. But that response is not linear — it follows a characteristic S-curve (the Hurter–Driffield curve) that compresses bright highlights and opens up shadow detail in a way that closely mirrors human vision. Film’s latitude — its ability to hold detail across a wide exposure range — was a built-in property of the medium. The toe of the curve gently lifts shadow detail; the shoulder rolls off highlights rather than clipping them.
Digital sensors: linear capture
A digital camera sensor works differently. Each photosite contains a photodiode that counts photons electrically. Double the light, double the electrical signal — the response is perfectly linear. This creates two practical problems.
First, the human visual system is not linear — we are far more sensitive to changes in dark areas than bright ones. A linear encoding wastes most of its code values on highlights the eye can barely distinguish, while crushing subtle shadow gradations into a handful of values. Second, a sensor’s full dynamic range cannot fit into a standard broadcast signal. A modern cinema camera captures 14–17 stops; Rec.709 video can represent around six to eight stops before clipping. Encoding directly to Rec.709 throws away most of what the sensor captured.
Log encoding: the solution
Log encoding applies a mathematical compression to the sensor’s linear signal before recording. The transformation redistributes code values to weight shadow and midtone detail more heavily — replicating what film’s chemistry did naturally — and squeezes the full dynamic range of the sensor into the available recording space. The result looks flat and desaturated, but all the captured information is preserved for grading.
Each camera manufacturer uses their own log encoding: ARRI’s Log C, Sony’s S-Log3, Blackmagic’s BRAW. All solve the same problem; they differ in their mathematical curve and the colour gamut they record into.
Why gamut matters
Sensors also capture a wider range of colours than Rec.709 can represent — particularly in greens and cyans. So log footage is recorded into a wide-gamut colour space — S-Gamut3, ARRI Wide Gamut, Blackmagic Wide Gamut — that can hold the full captured colour volume without clipping. This is why colour management exists: to track each input colour space, maintain a wide working space during grading, and transform precisely to whatever each deliverable requires.
Working this way — preserving the full captured data through the grade and applying the display transform only at the final step — is what the industry calls a scene-referred workflow. It is the modern standard, and the principle that Sections 02 through 06 elaborate on.
02 — Colour spaces and gamuts
A colour space is defined by three parameters: its colour primaries (which specific red, green, and blue define the gamut boundaries), its white point (what neutral looks like — D65 for most standards), and its transfer function (how values are encoded — linear, gamma, or log). Two images can have identical RGB numbers and look completely different if they were encoded in different colour spaces.
The key colour spaces
BT.1886 defines only a transfer function (EOTF — electro-optical transfer function), not a colour space. It has no primaries of its own. It is the standardised EOTF for Rec.709 SDR displays, specifying a near-2.4 gamma response. This is a common point of confusion: Rec.709 defines the colour primaries and white point; BT.1886 defines how those code values should be displayed.
P3: the bridge between broadcast and Rec.2020
DCI-P3 was developed for digital cinema projection and is the required delivery format for all DCPs (Digital Cinema Packages). It uses the P3 primaries with a slightly warm DCI white point (approximately 6300 K) and a simple 2.6 gamma transfer function. Every theatrical release is mastered to DCI-P3.
Display P3 uses the same colour primaries as DCI-P3 but with the standard D65 white point and an sRGB-like transfer function. This is the colour space used on all modern Apple devices and as the SDR streaming reference for platforms including Apple TV+ and Disney+.
When platforms specify P3-D65 for HDR mastering, they mean the P3 primaries with a D65 white point combined with the PQ transfer function, delivered in a Rec.2100 container. This is distinct from Display P3, which is an SDR colour space. Netflix’s HDR mastering guidelines specify P3-D65 primaries with PQ — mastering monitors should be calibrated to P3-D65 for HDR grades.
Although Rec.2100 uses Rec.2020 colour primaries as its container, no display currently achieves full Rec.2020 reproduction. In practice, HDR content is mastered to P3-D65 within a Rec.2100 container — it gives meaningful wide-gamut headroom over Rec.709 and can be accurately monitored with current display technology. Full Rec.2020-wide mastering is effectively theoretical at present.
P3 covers approximately 26% more colour volume than Rec.709, with the expansion mostly in greens and reds. Rec.2020 covers approximately 67% more than Rec.709. On a P3-capable display, the difference from Rec.709 is clearly visible — particularly in saturated foliage, skies, and skin tones.
03 — Colour management
In an unmanaged workflow, the grading application treats every piece of footage as if it is already in the working colour space. The colourist manually applies LUTs at the right points in the pipeline — and if anything is applied out of order, the colour is wrong. In a colour-managed workflow, the application knows the colour space of every input and every output, and handles the transforms automatically using mathematically precise Colour Space Transforms.
Scene-referred vs display-referred
Colour management frameworks divide into two fundamental approaches. A display-referred workflow transforms footage into the target display space early — typically Rec.709 — and all grading happens within those display constraints. This was the traditional broadcast model: footage was log-decoded and displayed on a Rec.709 monitor, and the colourist worked within the limits of that space.
A scene-referred workflow keeps the image in a scene-linear or log working space as long as possible. The display transform — converting the working space to the target display — is the very last step. Both RCM and ACES are scene-referred frameworks. The principle is consistent: work in the largest, most neutral space available, and transform to the display only when you need to view or deliver. This preserves maximum creative flexibility and mathematical accuracy throughout the grade.
CSTs vs LUTs
A Colour Space Transform (CST) is a mathematical formula that converts every input colour to its correct output colour using the defined equations of both colour spaces. A LUT approximates the same transform by pre-computing results on a grid of sample values and interpolating between them. The distinction matters:
In modern grading, CSTs have largely replaced LUTs for all technical transforms — camera input, working space conversions, and display output. LUTs remain the right tool for creative looks, show LUTs, and on-set monitoring (see Section 05).
Three approaches to colour management
Resolve Color Management (RCM) and DaVinci Wide Gamut
Blackmagic Design’s own colour management system — Resolve Color Management has existed since version 12; the modern DaVinci Wide Gamut Intermediate (DWG/I) working space arrived in Resolve 17. DWG/I combines two components: DaVinci Wide Gamut (the colour gamut) and DaVinci Intermediate (the log encoding applied to it). When RCM is enabled in Resolve, the application automatically applies the correct input transform for each camera format on ingest and the selected output transform for the viewer and every deliverable — no OCIO configuration required.
RCM has become the dominant choice for productions graded entirely in Resolve. It is faster to configure than ACES, ships with manufacturer-certified input transforms for all supported cameras, and handles mixed-camera productions (ARRI + Sony + Blackmagic on the same timeline) automatically.
The CST node in Resolve
Even in unmanaged Resolve timelines, colourists can insert explicit CST nodes anywhere in the node graph to perform a precise transform between two colour spaces. This is how a colourist working without full RCM can still use mathematically correct transforms rather than LUTs — for example, converting S-Log3/S-Gamut3 footage to DaVinci Wide Gamut at the clip level using the camera manufacturer’s certified colour science, then reverting at the output for the deliverable colour space.
Choose one colour management approach — ACES, RCM, or direct LUT — and apply it consistently across the entire project. Mixing approaches within a timeline produces incorrect results and is one of the most common sources of colour pipeline errors in post-production.
DaVinci Wide Gamut is not the same as Blackmagic Wide Gamut. Blackmagic Wide Gamut is the native colour gamut of Blackmagic camera sensors — a camera-native gamut like ARRI Wide Gamut or Sony S-Gamut3. DaVinci Wide Gamut is a grading working space, designed to contain all camera gamuts without clipping. Both appear in a Resolve RCM workflow at different stages of the pipeline.
04 — Log formats
Every major camera manufacturer ships their own log encoding and associated wide-gamut colour space. In a colour-managed pipeline, the colourist sets the input colour space per clip (or globally per camera) and the system handles the transform automatically. Understanding log formats is still necessary — you need to know which format a clip was recorded in to declare it correctly. A wrong input declaration produces subtly or obviously wrong colour throughout the grade.
Log footage is not broken footage. It is footage encoded to preserve maximum tonal information for grading. Viewing log material without a display transform on a standard monitor gives you no useful information about how it will look in the finished grade.
Always confirm the camera’s log format and firmware version with the DIT or camera department before starting a grade. Log C3 and Log C4 are not interchangeable — applying a Log C3 input CST to ALEXA 35 footage (Log C4) produces incorrect colour and clipped highlights.
05 — LUTs: still essential
LUTs have not disappeared from modern post-production — they have found their natural role. As technical transforms have moved to mathematically precise CSTs, LUTs remain the right tool for the jobs that require a baked, portable, real-time colour mapping: on-set monitoring, creative looks, show LUTs for dailies, and output where a CST is not available.
1D vs 3D LUTs
Grid size and precision
Where LUTs still belong
On-set monitoring. A camera assistant’s monitor or a director’s reference display cannot run a full colour-managed pipeline in real time. A show LUT — baked to a 17- or 33-point cube for the specific camera format and display — is applied to give on-set viewers a consistent representation of the intended look.
Dailies and editorial. The show LUT distributed to editorial ensures that editors and producers see something close to the intended grade on offline monitors, without needing a colour-managed system.
Creative and show LUTs. The visual identity of a production — a drama’s warmth, a thriller’s cool desaturation, a period film’s emulsion-like contrast — can be encoded as a show LUT. Colourists increasingly build these looks as CST-based grades and then export the result as a LUT for distribution to set and editorial, giving the portability of a LUT file while retaining the precision of a managed grade in the suite.
Film emulation. LUTs for film stock emulation (Kodak 2383 print stock, Fuji 3513, etc.) remain common as creative starting points. These cannot be replicated accurately by a CST — they encode specific aesthetic choices about grain response, colour cross-over, and saturation behaviour that are creative rather than technical.
In a colour-managed pipeline, LUTs are applied within the working colour space — after the input CST and before the output CST. A creative LUT designed for Rec.709 must not be applied to log footage or inside a wide-gamut working space. Applying LUTs at the wrong point in the pipeline is the most common source of unexpected colour casts and contrast shifts.
06 — ACES
ACES — Academy Color Encoding System — is an open colour management framework developed by the Academy of Motion Picture Arts and Sciences. It defines a complete pipeline: a scene-linear reference space large enough to contain all real colours, standardised camera input transforms, and output transforms for every delivery standard. ACES and Resolve RCM are different implementations of the same principle (see Section 03); ACES’s key advantage is that it is facility-agnostic and supported across all major grading applications. ACES 2.0, released in late 2024, is now the current version — shipping in OCIO 2.4, DaVinci Resolve 20, Autodesk Flame, and Pomfort Livegrade.
The ACES colour spaces
Input Transforms (IDTs) and Output Transforms (ODTs)
Getting footage into ACES requires an Input Device Transform (IDT) — a camera-specific conversion from the camera’s log colour space into the ACES reference space. IDTs are provided by camera manufacturers and maintained by the Academy. An ACES IDT is designed to produce scene-linear, perceptually accurate results in the reference space, not merely to look correct on a Rec.709 display.
Getting the grade out of ACES requires an Output Device Transform (ODT) — a conversion from the working colour space to a specific display standard. ACES provides ODTs for Rec.709, DCI P3, Rec.2020 PQ (HDR10), Display P3, and Dolby Vision. The same grade produces correct results for any display by swapping the ODT — the core value of ACES for productions with multiple delivery formats.
When to use ACES
For simpler productions — a single camera, a straightforward Rec.709 delivery, all work in Resolve — Resolve RCM is usually faster and equally accurate. ACES adds pipeline complexity. That complexity is worth it when the benefits above apply; it is unnecessary overhead when they do not.
ACES 1.3 vs ACES 2.0
ACES 2.0 redesigns the Display Rendering Transform — the perceptual rendering step at the end of the ACES pipeline. The principal failure of ACES 1’s RRT (Reference Rendering Transform) was hue instability under narrow-band emitters: LEDs and lasers produced objectionable colour shifts in saturated reds and greens that are common in music video, commercial, and performance footage. ACES 2.0 addresses this with a gentler tone scale, reduced mid-tone contrast, and substantially improved hue stability. The RRT and ODT are unified into a single, more configurable Display Rendering Transform.
ACES 2.0 also resolves the roundtrip limitation that complicated VFX work and SDR/HDR cross-mastering in ACES 1: Output → ACES → Output conversions now work cleanly. ACES 1’s RRT was effectively one-way. Reference Gamut Compression (RGC) — introduced in ACES 1.3 to handle out-of-gamut values from saturated practical lights — is generally disabled in ACES 2.0 pipelines, because the new DRT handles gamut compression natively. The ACES Metadata File (AMF) specification has also been updated for improved cross-facility interchange. In September 2025, ACES became an Academy Software Foundation project, aligning it more tightly with OpenColorIO development.
For new projects, ACES 2.0 is the version to choose unless a facility partnership specifically requires 1.3. It is available in project settings wherever ACES is supported — look for ACES 2.0 or OCIO 2.4 in the application’s colour management preferences.
Custom DRTs
Beyond the standard ACES DRTs and Resolve’s built-in output transforms, a growing category of custom Display Rendering Transforms has emerged from colourists and developers. Cullen Kelly’s Referent, the FilmVerse family, Open DRT, TCAM, and AgX each implement different aesthetic approaches to the scene-to-display rendering step — prioritising hue stability, smoother highlight rolloff, or a more filmic response. DaVinci Resolve 20 added explicit support for custom DRTs within the RCM framework, making them accessible without requiring a full OCIO pipeline.
07 — HDR and P3 delivery
High Dynamic Range video extends both the brightness range and colour volume of an image beyond what a standard Rec.709 signal can carry. A standard SDR display targets around 100 nits peak and Rec.709 colour. Modern HDR displays target 1000 nits and above — reference monitors such as the Sony BVM-HX3110 and Dolby Pulsar reach 4,000 nits peak and reproduce a significantly wider gamut — typically targeting P3-D65, with Rec.2100 (Rec.2020 primaries) as the container. Three HDR formats are relevant to delivery, alongside the P3 colour space requirements for cinema and streaming.
HLG — Hybrid Log-Gamma
HLG was developed jointly by the BBC and NHK for broadcast. Its key property is backwards compatibility: an HLG signal displayed on a standard SDR television without any conversion looks like a normal SDR broadcast picture. This makes it practical for live transmission, where a single signal must serve both HDR and SDR receivers simultaneously. HLG carries no metadata — the signal contains all the information needed for display, simplifying it for broadcast chains.
HDR10
HDR10 uses the PQ (Perceptual Quantizer) transfer function, encoding absolute luminance values — a code value always represents the same number of nits regardless of the display. HDR10 carries static metadata (MaxCLL and MaxFALL — maximum content light level and maximum frame-average light level) which the display uses for tone mapping. HDR10 is an open standard and is the baseline HDR format required by every streaming platform. Content is delivered in a Rec.2100 container (Rec.2020 primaries), typically mastered at P3-D65.
HDR10+
HDR10+ extends the HDR10 standard with dynamic metadata — per-scene brightness information that allows the display to tone-map each scene individually, similar to Dolby Vision. Unlike Dolby Vision, HDR10+ is an open, royalty-free format requiring no licensing agreement or Dolby-certified facility. Netflix began streaming HDR10+ in 2025 alongside HDR10 and Dolby Vision. In practice, Netflix derives HDR10, HDR10+, and SDR versions from a submitted Dolby Vision master — they do not accept separate HDR10+ deliverables. The HDR10+ dynamic metadata is generated by Netflix from the Dolby Vision trim pass.
Dolby Vision
Dolby Vision is built on the same PQ transfer function as HDR10 but adds dynamic metadata — brightness and colour information that changes scene by scene. Where HDR10 gives the display a single set of numbers for the entire programme, Dolby Vision gives it per-scene instruction data, allowing more precise tone mapping for each scene. On a Dolby Vision-capable display this typically produces more accurate shadow detail and highlight rendering than HDR10 with static metadata.
Dolby Vision requires a licensed Dolby-certified facility for the trim pass — a secondary adjustment that creates the SDR version from the Dolby Vision master. The trim pass metadata travels in the file and allows a single deliverable to serve Dolby Vision, HDR10, and SDR playback.
Netflix accepts only Dolby Vision masters as their ingest format. HDR10, HDR10+, and SDR versions are derived by Netflix from the submitted Dolby Vision master. Facilities do not need to deliver separate HDR10 or HDR10+ files for Netflix.
P3 in cinema and streaming delivery
Cinema DCP requires DCI-P3 — the full P3 gamut with the DCI white point (approximately 6300 K, slightly warm relative to D65) and a 2.6 gamma transfer function. Cinema projectors are calibrated to this standard. A DCP graded on a D65-calibrated monitor must account for the white point difference to avoid a warm colour cast in the cinema.
Streaming platforms use Display P3 (P3 primaries, D65 white point) as their wide-gamut reference. Netflix requires P3-D65 as the SDR reference display for HDR mastering. Apple TV+ delivers content in Display P3 on P3-capable Apple devices. Disney+ uses Display P3 for its HDR reference. In practice, a properly mastered HDR10 deliverable (PQ, Rec.2100 container, P3-D65 mastering) satisfies the requirements of all major streaming platforms.
Simultaneous SDR and HDR delivery
Most streaming platforms require both an HDR master and an SDR version. The two are not brightness-adjusted copies of each other — HDR and SDR represent different creative decisions about how the image should look on their respective displays. Highlights that are rendered with full detail in HDR may need to be compressed differently for SDR; colours that exist in P3 need to be mapped back within Rec.709.
The industry-standard approach is to grade the HDR version first in a colour-managed pipeline, then create an SDR trim pass. In a Dolby Vision workflow the trim pass is embedded in the Dolby Vision metadata; in a non-Dolby workflow it is a separate SDR grade created by the colourist from the HDR reference. Either way, the SDR and HDR grades share the same creative intent — they are not independent grades.
Plan HDR delivery into the post schedule from the start. An HDR grade in a colour-managed pipeline takes more time than a standard SDR grade, and the trim pass requires a separate session. Treating HDR as an afterthought — requesting it after an SDR grade is locked — often means rebuilding the grade from scratch.
Resolve 20.2 note: DaVinci Resolve 20.2 (September 2025) updated RCM and CSTs to use ITU BT.2408 for HLG ↔ PQ conversion. Productions using HLG or cross-converting between HLG and PQ should ensure they are on Resolve 20.2 or later for accurate results.