Digital product colour: what an AI/CGI colour workflow can establish

Prepared by SHAR Production ·

Desk review of public primary sources.

A red product can look different in a catalogue image, a neutral studio scene and a warm advertising environment. The commercial problem is not necessarily the difference. It is the inability to explain whether it comes from the material, lighting, creative treatment or interpretation of the file.

Hybrid AI/CGI production makes that question tangible. A composition may combine a digital model, approved packaging artwork, a generated environment and graphic overlays. Buyers need a way to agree how these elements work together while protecting product identity. The name of a colour tool in a proposal cannot make that agreement on its own.

Research question and method

This desk study asks what evidence a buyer should request to distinguish a controlled representation of colour from a general promise that the colour will be correct.

The selected sources are official ACES documentation on encodings and display encoding, together with the official OpenColorIO website. They were checked on 27 September 2026. Selection required a direct description of the component's role rather than a studio marketing example. No software speed comparison, display measurement or generative model experiment was conducted.

Documented findings appear first. The purchasing and acceptance framework that follows is our original synthesis. It translates technical distinctions into project decisions rather than proposing an industry certification scheme.

Three different foundations for discussing colour

ACES documentation distinguishes the system from specific encodings, including ACES2065-1, ACEScg and ACEScct, which serve different workflow purposes. The page is dated 10 September 2025.

A buyer therefore needs more information than the word ACES to understand the path taken by a particular file. ACES encodings.

The Display Encoding documentation describes preparation for a display, taking account of its configured colour gamut and signal transformation. That page is also dated 10 September 2025.

The useful purchasing implication is to connect image approval with stated viewing conditions. ACES Display Encoding.

OpenColorIO describes its role as colour management for motion pictures, visual effects and computer animation.

This identifies a workflow tool; it does not certify the match between a particular product image and its reference. The checked homepage provides no update date. OpenColorIO.

These sources address different questions. An encoding describes image data. A display transformation describes presentation. A tool supports processing. Deciding whether a depicted shade is acceptable for a particular brand remains a separate approval decision.

Establish a reference hierarchy before discussing corrections

A folder of product photographs can contain conflicting instructions. One image shows a saturated body; another shows a pale version of the same object. Treating either as an unquestionable reference can start a long correction loop without resolving what the buyer actually wants.

A useful brief assigns roles. Approved packaging artwork governs graphics. An agreed control image governs material appearance under a stated lighting condition. Campaign references govern mood. Where they conflict, the supplier should know which has authority. Older images with an unknown processing history should be identified as inspiration where that is their actual role.

The brief should also describe relationships between variants. A product family may need visibly distinct shades even when each individual image looks attractive. If two versions become difficult to distinguish in the final comparison, a technically orderly export has still missed an important commercial requirement.

A buyer's evidence matrix

Sources are interpreted consistently. Evidence to request: Source inventory and agreed settings. What remains outside that evidence: Accuracy of the sources themselves.

A control appearance is approved. Evidence to request: Approved frame and viewing conditions. What remains outside that evidence: Appearance in every later scene.

Delivery is reproducible. Evidence to request: Version identifier, output settings and control file. What remains outside that evidence: Identical presentation on every device.

Product identity is retained. Evidence to request: Critical regions compared with references. What remains outside that evidence: Accuracy of unreviewed product variants.

Placement is accepted. Evidence to request: Review of the actual uploaded version. What remains outside that evidence: Behaviour on every other platform.

Agreeing this matrix early makes responsibilities visible. Some questions require a technical specialist; others need a brand owner's decision. Without that separation, colour review can become an exchange of screenshots that have no common reference.

The matrix also helps purchasing teams understand two superficially similar quotations. One may include an approved control stage and delivery review, while the other may include only a finished export. Those are different scopes even if their headline duration and resolution match.

Separate environmental freedom from product control

For a product film, define which elements can change creatively. The mood of a room, cloud formations or decorative surroundings can usually be discussed as art direction. Body colour, label graphics and differences between product variants need their own acceptance requirements.

This leads to a practical choice. If a generated scene is compelling but the product cannot reliably retain its approved appearance, consider a separately controlled product layer. If the object is small and its exact shade is not carrying the message, a lighter review scope may be sufficient. The scene's purpose should determine the decision.

Check the difficult lighting context before commissioning the entire sequence. A product that looks convincing against a neutral background may still need evaluation under strong coloured light. A focused visual test can make that decision reviewable before it affects many finished shots.

Hypothetical example: two matte bottle variants

Consider a campaign for plum and burgundy bottles. Each appears separately in a warm interior; both appear together in the final shot. Close-ups must also communicate the matte surface.

Start by agreeing a neutral control image of the pair. Then decide how much convergence between the shades is acceptable in the warm scene. Finally, preserve a clear comparative view where the audience can distinguish the variants. The discussion now concerns a specific viewing result rather than a request to make the red more correct.

An acceptance note might state: the variants remain distinguishable in the final paired frame; strong lighting does not reverse their perceived relationship; the label remains readable at the agreed presentation size. These are project-specific visual criteria. A numerical colour tolerance would require a separately defined measurement procedure, samples and conditions.

The buyer should also appoint one person to consolidate colour feedback. Separate comments from packaging, marketing and sales can otherwise describe different objectives as though they were one correction. Consolidation makes the disagreement visible before another render is commissioned.

Approve the exported asset as a separate deliverable

Review in a working application and review of the delivered file are separate project activities. Record the exact accepted version, the purpose of the file and the agreed review method. A folder link without a version identifier makes it harder to establish what was approved.

If placement review is included, compare the actual uploaded asset with the supplied deliverable in the selected viewing scenario. When a difference appears, investigate where it entered the chain. Recolouring the original without that diagnosis may compensate for one problem while creating another elsewhere.

A messaging screenshot can identify a region that needs attention. Final approval should return to the agreed file and viewing arrangement. Useful feedback names the version, timecode, region and observed difference. It also distinguishes an unexpected change from a new creative preference.

Information needed for a useful quotation

For a CGI product film, provide reference assets, product variants, mandatory scenes and the intended uses of the final files. Flag colours that communicate function: distinguishing a SKU, indicating device status or identifying a component in an explanatory view.

Ask the proposal to identify the control stage before extensive rendering, the person responsible for colour approval and the documentation included at handover. If the materials will be reused later, agree which settings and source assets will accompany them. This makes the scope comparable across proposals.

Limits of the study

This review does not establish numerical colour tolerances or verify a particular production. Documentation cannot demonstrate error-free operation, accurate source photography or identical appearance on all screens. Physical print matching and metrological procedures were outside the selected scope.

Purchasing conclusion

The useful deliverable is a reproducible sequence of decisions: reference, control appearance, processing conditions, output purpose and acceptance of the actual result. A colour-management system becomes meaningful to the buyer when it supports that sequence.

Attach conflicting references to the project brief and explain which product differences must survive. This gives the production team a concrete problem to solve while leaving room for deliberate art direction.

Sources and methodology

Author and creator: SHAR Production. Published: . Modified: .

Related reading: Agreeing CGI product light and material through comparable samples.

For commercial CGI, approve colour and material references before the final render and identify the shots and delivery versions included in production. 3D product video production.