FRAMES
What Images Are Not
How visual truth was manufactured long before generative tools

▲ Fictional generative image created for this article,
reconstructing Doisneau’s Parisian kiss scene as a staged setup,
with the photographer directing the couple and nearby extras.
▫️ The present alarm over synthetic images depends on a convenient misreading. Generative tools did not create the image lie. They exposed an older arrangement in which realism is routinely granted the force of proof. The system is new; the settlement is not. Images keep acquiring the authority of established fact because that authority reduces the labour of interpretation, allowing viewers to treat the surface as though judgement had already been completed.
The arrangement begins with operations so ordinary they are rarely named. The first manipulation is the frame; in moving images, it is also the cut. Both subtract while presenting themselves as clarification. They decide what will count as the event, what will recede into background, and what will disappear before it can register as evidence. Exclusion rarely appears as exclusion. More often it arrives as irrelevance. The issue, then, is not simply that images can be fabricated. It is that visual trust has long been manufactured through formal decisions, institutional tests, and circulation systems that make a disciplined surface look like an adequate account.
Under those conditions, the recent anxiety around synthetic images changes scale. Fabrication was never the novelty. What generative systems add is efficiency: they compress the labour required to produce cues viewers have already been trained to trust. Continuity reads as honesty, grit as sincerity, high production values as authority. The pattern predates the tool. The tool industrializes it.
Art culture offers one of the clearest sites for seeing this because it does not pretend images are unmade and yet still permits them to circulate as evidence. That double status matters. It allows style, persuasion, and documentary authority to inhabit the same surface without obvious conflict. Photography and cinema occupy that overlap with unusual force. The same object can hang in a gallery and move through the world as testimony; the same clip can be screened as cinema and then used as “what really happened”.
Robert Doisneau’s *Le baiser de l’Hôtel de Ville* remains useful because it clarifies the terms of belief. Images do not need factual spontaneity to secure social trust. They need only satisfy a learned image of spontaneity. The posed nature of the scene became widely known, but its symbolic authority endured. Once an image has met a prior expectation of what “real life” should look like, exposure does not necessarily revoke its credibility.
The question, then, is larger than one photographer’s ethics. It concerns the grammar through which images persuade. Belief does not arrive through metaphysical certainty so much as through cues and genre conventions that viewers have been trained to register as atmosphere rather than argument. A certain framing conveys intimacy; a certain lighting confers seriousness; a certain grain implies unvarnished access. Once those cues land, the image begins to function socially as reality whether or not it deserves that standing.
Beauty intensifies the pattern because it alters the terms on which suffering is received. Ingrid Sischy’s critique of Sebastião Salgado matters here not as a final ruling on motive, but as an account of what formal mastery can do once it absorbs ethical resistance. Beauty can convert distress into a completed experience for the spectator. What should remain abrasive becomes legible, moving, even satisfying. The viewer receives the pleasures of seriousness while being shielded from implication. Tonal control and printing density help determine whether pain appears contingent and intolerable or monumental and therefore strangely settled. At that point, craft is no longer merely styling the image; it is preparing the image to circulate as witness.
Technical language belongs at the centre of the argument for the same reason. The lens organizes distance: a wide angle coerces intimacy, while a long lens turns people into specimens and presses the world flat. Angle assigns rank, focus allocates attention, and light issues judgement. Colour participates in the same ordering. The familiar blue-yellow treatment matters not because it expresses a timeless symbolism, but because it has hardened into an editorial habit with geopolitical consequence. Cool blues often help frame the United States as composed, competent, systemically secure; hot yellows often help frame Mexico and Latin America as hot, dusty, and ambiently disordered. Encountered once, this can read as style. Encountered across films, news packages, travel reels, and platform video, it begins to pass as description. The palette does not simply render a place. It distributes value and lets that distribution pass as realism. In that regime, “raw” is less a truth claim than an authorized texture.
Reading Time
11 MINUTES
Words
GIULIANO CESAR
Header
G—G LAB
Published
MAR 2026
Reading Time
11 MINUTES
Words
GIULIANO CESAR
Header
G—G LAB
Published
MAR 2026
The first manipulation
is the frame.
In moving images,
it is also the cut.

In Traffic (2000), the stark yellow and blue grading drew criticism for reinforcing stereotypes, casting Mexico as “dirty” and overheated and the U.S. as clinical and cold, a recurrent Hollywood shorthand that reduces places to coded visual atmospheres.
▲
The first manipulation
is the frame.
In moving images,
it is also the cut.
Moving images extend the same logic because they govern not only what is visible but the pace at which interpretation becomes possible. Riefenstahl remains the clearest demonstration: spectacle staged for the lens, vantage and scale engineered to assign rank, sequences built so that power appears not asserted but ordained.
Editing is where arrangement hardens into meaning. Coincidence becomes plot, proximity becomes motive, and rhythm allocates moral weight before reflection can fully catch up. Duration prepares affect in advance of judgement. Continuity feels natural only because cinema trained viewers to experience construction as flow, and that engineered flow still circulates as reality.
Sound continues the same work by placing instruction inside atmosphere. Voiceover resolves ambiguity into certainty. Music tells viewers what kind of feeling the scene deserves. Subtitles often clean speech into firmer claims than the speaker made, and silence can be just as directive: a muted chant is not absence but management. Video is treated as superior evidence because it appears to preserve time. What it preserves is selected time.
Nothing here requires synthetic generation. It requires choices, and those choices arrive attached to incentives and constraints. The camera is rarely a solitary eye. It is labour inside an economy of images shaped by access, budgets, risk, institutional permission, editorial taste, and distribution demand. Aesthetic construction becomes political because those pressures determine what can be shown and what must be made legible enough to circulate. Clarity has a discipline of its own. It prefers pictures that resolve quickly, not innocently, but because quick resolution reduces the burden of interpretation and therefore carries institutional value. What circulates as evidence, however, is not decided by the image alone. It is decided again by review, eligibility, and admissibility: by the systems that convert a record into documentary authority.
That is where institutions enter, not as afterthoughts but as administrators of credibility. Public argument often behaves as though truth clings to an image on its own. Institutions do not. They build rulebooks, checks, and review systems because they know that the line between acceptable adjustment and altered meaning remains under permanent pressure from incentives to dramatize, beautify, and simplify.
World Press Photo makes the point with unusual clarity. Its contest rules draw a bright line: photographs must be made with a camera; synthetic or artificially generated images are not permitted; generative fill is prohibited. Around those prohibitions sits a review apparatus that defines what counts as manipulation and how authenticity will be tested. This does not uncover an essence called truth. It establishes the basis on which an image may claim documentary authority.
The recent history of those rules matters because it shows legitimacy being revised in public. In late 2023, World Press Photo signalled that generative tools would be permitted in an Open Format category, then amended the guidance to ban them there as well. The significance lies less in policy confusion than in institutional disclosure. Verifying a file is one task; granting it documentary standing is another. Authenticity at the level of production is one test; institutional legitimacy is another. The “real image” that enters public circulation is therefore not simply the one that exists, but the one that survives admissibility criteria with authority intact.
Even iconic images can fail to remain stable under those grounds. In 2025, controversy over the authorship of the “Napalm Girl” photograph led World Press Photo to retain the award while suspending attribution because doubts could not be settled to its satisfaction. The episode is revealing because the problem exceeds pixels. Attribution, archive, caption, and institutional need all help determine what an image is allowed to be. Documentary truth concerns not only whether the scene existed, but who is authorized as witness and how that authorization is sustained.
Images do not need
factual spontaneity
to secure social trust.
They need to satisfy a learned
image of spontaneity.

Triumph of the Will (1935), directed by Leni Riefenstahl,
remains one of cinema’s most effective and infamous propaganda films.
▲
Images do not need
factual spontaneity
to secure social trust.
They need to satisfy a learned
image of spontaneity.
Once credibility is understood as administered, documentary authority starts to look less like an intrinsic property than a power position that can be lent or withheld. It attaches to images that survive review and align with institutional needs, and it fails to attach to images that threaten those needs. Many images persuade because they are easy to read, because they fit an existing narrative, because they arrive already prepared for circulation. Circulation is not a reward for truth. It is often a reward for usefulness.
Soft power is usefulness naturalized as common sense. Catherine Lutz and Jane Collins’s *Reading National Geographic* remains instructive because it shows editorial ordering disappearing into repetition. A magazine widely received as neutral education produced a durable picture of “the world” through decisions about subjects, captions, and typologies of personhood. Some places appeared modern; others appeared outside time. Repetition converted editorial ordering into apparent reality.
A comforting distinction usually enters here. These are said to be matters of taste and bias, while true manipulation belongs to propaganda states, elsewhere and earlier. The distinction does not survive inspection. Soviet image censorship, with purged figures removed and the visible world cleaned to match state narrative, makes the same structure visible at maximum institutional force. Memory is administered by editing the record.
That case matters because it clarifies the structure. Reality control does not depend on fabrication alone. It also depends on erasure and non-circulation. The public record is not the archive. It is the portion allowed to become visible, repeatable, and credible.
From there, the contemporary variant follows directly. The mistake is to reserve the concept for blunt falsification when power also manages reality by governing visibility in time. Release can do the work of deletion. Partial disclosure, contested access, redaction folded into the performance of transparency: all of these can produce conviction without closure.
People continue to cling to the belief that truth is secured at capture, that the camera’s presence guarantees what matters. It does not. Public truth is manufactured downstream. An archive can be vast and public reality still narrow if distribution prevents most of that archive from becoming visible. In war, invisibility is not absence. It is policy.
What circulates as evidence
is not decided by the image alone.

In February 1982, National Geographic digitally moved the Pyramids of Giza
to fit a vertical cover, a watershed moment in debates on image manipulation.
▲
What circulates as evidence
is not decided by the image alone.
Gaza has become a concentrated case because the sequence is unusually plain. Access restriction and platform filtering are not separate problems so much as consecutive stages in the manufacture of public reality. The first stage is capture. Through access regimes framed as security and routine administration, independent witnessing is narrowed before the record is even made. Since 7 October 2023, Israel has barred international journalists from entering Gaza independently, allowing only limited, military-escorted visits in some instances. The matter has proceeded through the Israeli Supreme Court, while press freedom organizations have condemned the restriction as incompatible with legal commitments.
The consequence follows immediately. Restrict independent access and the record contracts at the point of capture. What remains visible is what authorized vantage permits. The official version prevails first by controlling who gets to see.
The second stage is circulation. Platforms sort visibility after the fact. Moderation and ranking determine which fragments become the dominant picture, and Human Rights Watch has described patterns of censorship of Palestine-related content on Instagram and Facebook that shape what remains publicly legible. First a narrower record is produced; then that narrowed record is filtered into common reality.
Synthetic images and deepfaked video are dangerous because they scale plausible fabrication and compress the cost of manufacturing cues audiences have long been trained to trust. The danger is real. States will exploit it. Opportunists will exploit it. So will boredom and cruelty, because outrage is now a standing revenue model and plausibility is a cheap weapon.
For that reason, the crisis cannot be treated as purely technical, as though detection alone could restore an earlier innocence. Detection matters. Provenance matters. Forensics matter. None of them replaces the larger shift now required. The relevant question is no longer simply whether an image is real, as though reality were a stamp applied to the surface, but what this realism hides, what incentives produced it, what checks certify it, what distribution systems elevate it, and what alternatives were prevented from circulating long enough to enter public knowledge.
The value of that shift is responsibility, not cynicism. It does not require abandoning images. It requires giving up the fantasy of the image as closed proof. Framing, craft, certification, and circulation are not secondary complications. They are the sites at which credibility is made.
Generative tools did not break visual truth. They interrupted the habit of locating truth in the image alone. That habit was the older illusion. What now feels like crisis is partly the loss of a convenience: the convenience of letting surfaces carry judgement. The panic is recent. The arrangement it exposes is not. ◾️



