Should Historical Footage Be Colourised? A Decision Tree for Film Societies

Begin With the Purpose of the Request

Requests to colourise historical footage often arrive with a straightforward ambition: make an old film easier to watch, create a more immediate connection with younger audiences or prepare material for an exhibition. A film society may also receive requests from families, local organisations or documentary producers who want monochrome footage to appear closer to contemporary moving images. These are different objectives and should not automatically lead to the same intervention.

Colourisation changes the interpretative character of a historical image. A monochrome film records differences in light and tone, not a complete map of original colours. A colourisation system estimates how objects might have appeared using visual patterns, contextual information and learned associations. Its output can be persuasive without being historically established. The first question is therefore what the colourised version is expected to communicate.

• Is the purpose accessibility, interpretation, exhibition or entertainment?
• Will the colourised version be presented as historical evidence?
• Is there reliable information about the original colours?
• Can the society preserve and disclose the source version?

Decision Point One: Establish the Source

Identify the original carrier and establish whether the available material is a monochrome camera original, a later duplicate, a scanned print or an existing digital transfer. The distinction matters because the source may already have undergone tonal adjustments, contrast changes or other interventions. Record its known provenance, approximate date, location, production context and any existing documentation. If these details remain uncertain, state the uncertainty in the catalogue.

The condition of the source also affects the decision. Heavy scratches, faded contrast, unstable exposure and missing frames may make colour assignment difficult. A restoration workflow should distinguish image repair from colourisation so that a model does not conceal physical damage by generating new surface detail. Where an existing scan is poor, a better preservation capture may be more appropriate than adding colour to an unreliable image.

For example, a local film society receives a five-minute monochrome sequence showing a street procession. The footage includes shopfronts, clothing, vehicles and banners. The society knows the approximate town but has no reliable record of the original colours. The first task is to establish what can be documented about the footage, not to select colours that make the sequence appear contemporary.

Decision Point Two: Assess the Evidence for Colour

Colour evidence exists on a spectrum. Original colour elements in surviving production material offer a different level of support from written descriptions or visual conventions associated with a historical period. A surviving colour photograph of the same street may provide useful contextual information, but it does not establish the colour of every object in the moving image. A uniform described in a contemporary document may support an interpretation of the clothing, although individual garments could differ.

An evidence record should distinguish direct evidence, contextual evidence and model inference. Direct evidence may include surviving colour frames from the same production or identifiable objects documented in contemporary colour photographs. Contextual evidence may include catalogues, municipal records and reliable descriptions. Model inference includes colours assigned because they are statistically or visually plausible. These categories should not be combined into one general claim that the footage has been restored to its original colours.

• Direct evidence: surviving original colour material or reliable matching references.
• Contextual evidence: contemporary descriptions, photographs and documented objects.
• Partial evidence: information supporting some objects but not the complete frame.
• Model inference: colours estimated without direct confirmation.

Decision Point Three: Choose the Appropriate Route

If the original colour can be established for the relevant material, consider whether a carefully documented colour reconstruction serves the intended purpose. The reconstruction should identify its references and explain any remaining uncertainty. If evidence covers only selected objects, avoid extending those findings to unrelated areas of the frame.

If evidence is partial, a limited colourisation may be appropriate for an interpretative presentation, provided the uncertain areas are clearly identified. For example, the colour of a documented municipal vehicle might be supported by a contemporary reference, while the clothing of spectators remains uncertain. The catalogue should describe these different levels of confidence instead of presenting the entire image as equally reliable.

If the available evidence is insufficient, consider retaining monochrome footage as the principal access version. A separately labelled experimental colourisation may still be useful for an educational discussion about historical interpretation, but its status should be unmistakable. If the purpose is simply to improve viewing, tonal correction, stabilisation or careful cleaning may address the request without introducing estimated colour.

• Strong evidence: proceed with a documented, evidence-led reconstruction.
• Partial evidence: restrict claims to supported objects and disclose estimates.
• Weak evidence: retain monochrome as the principal presentation.
• Unclear purpose: clarify the intended use before authorising intervention.

Decision Point Four: Evaluate the AI Intervention

AI colourisation systems estimate colour relationships from image content. They may identify a scene as an outdoor street, distinguish apparent skin from clothing and assign plausible colour ranges to different objects. However, the model cannot independently establish the actual colour of a particular coat, banner or building from a monochrome frame. Similar tonal values can represent very different colours, and historical materials may have had unusual finishes or appearances.

Automated systems can also produce inconsistencies between frames. A red-looking garment may become brown in the next shot, while a sign can change colour as the camera moves. Objects crossing the frame may receive unstable colour boundaries. Manual correction can improve visual consistency, but a human adjustment is still an interpretative choice unless supported by historical evidence.

A society should identify the model or service where known, document its version and settings where available, and distinguish automated assignments from subsequent manual corrections. Record the sections processed and retain the monochrome source. If the tool cannot provide adequate information about its operation, the catalogue should describe that limitation without attributing unsupported technical capabilities to it.

Decision Point Five: Approve the Presentation and Disclosure

The final decision concerns how audiences will encounter the footage. A colourised version should be labelled as a colourised derivative, with a concise explanation of the evidence and inference involved. Avoid titles or captions that imply the original appearance has been recovered when the colours were estimated. If the footage is displayed in an exhibition, the explanation should be visible beside the material rather than buried in a technical document.

A useful caption might read: This monochrome historical film has been digitally colourised for interpretation. Some colour assignments are informed by surviving contextual evidence; other details are estimated by an automated system. The original monochrome version remains available. This wording distinguishes the historical source from the modern intervention without requiring visitors to understand the technical operation of the model.

Preserve the original digital scan, the colourised derivative and the associated documentation as separate collection objects or clearly linked files. Where the society has limited storage, prioritise the preservation source and the final derivative, together with essential processing records. Any intermediate files retained should have clear identifiers and a defined relationship to the final output.

The Decision Tree in Practice

The following decision sequence can be applied to a new colourisation request. It does not assume that every request deserves a colourised result. Its purpose is to make the reasoning explicit and to give societies a consistent way to explain decisions to donors, audiences and researchers.

• Identify the original material and establish its provenance.
• Determine the purpose of the requested colourisation.
• Locate direct and contextual evidence for original colours.
• Separate documented colour information from estimated assignments.
• Choose evidence-led colourisation, limited interpretation or monochrome presentation.
• Preserve the original, label the derivative and publish an accessible disclosure.

A Worked Example: The Local Street Procession

Consider the hypothetical five-minute procession. The society identifies the town and finds a contemporary colour photograph of the municipal vehicle appearing in the film. This supports a colour reference for that vehicle, although differences between individual vehicles and recording dates still require consideration. No equivalent evidence exists for the spectators' clothing or the shopfront decorations.

The society can now make a proportionate decision. It might create a limited interpretative colourisation, using the photograph as a reference for the identifiable vehicle and disclosing that other colours are estimates. Alternatively, it might retain the monochrome footage as the main version and use selected colourised frames in an exhibition about the uncertainty of historical reconstruction. Either approach is more informative than presenting a fully colourised sequence as an authenticated reconstruction.

The next practical step is to write a short evidence inventory for the requested footage, separating documented colours from contextual clues and unsupported estimates. This record should exist before any colourisation begins and should determine the wording used in the final catalogue entry.

Questions readers ask