Design Is Not a Button

Published On: June 28th, 2026Last Updated: September 14th, 20265798 words29 min readDaily Views: 2Total Views: 11

A button can produce an image.
It cannot decide whether the image is right.

That distinction is becoming harder to see because generative systems can produce polished visual material almost instantly. A person enters a sentence, waits a few seconds, and receives something that appears finished:

  • the lighting is dramatic;
  • the colours are coordinated;
  • the subjects are rendered clearly;
  • the environment looks expensive;
  • the image resembles an advertisement, editorial photograph, book cover, or cinematic still.

To someone looking only at the final surface, it may seem as though the design work has already happened.

The image exists.

What else could be left to do?

Quite a lot.

An image can be technically impressive and conceptually wrong.

  • It can be beautiful but unsuitable for the audience.
  • It can follow the prompt while failing the purpose.
  • It can contain the correct logo but misrepresent the organisation.
  • It can show every requested person while communicating the wrong relationship among them.
  • It can look like a finished campaign asset while lacking the space, hierarchy, accuracy, accessibility, or consistency required for actual publication.

This is the problem hidden by the Generate button.
The button collapses production time.

It does not collapse the need for judgment.

The Visual Register by Mithaq Praxis is partly being built to make that distinction visible again.

Generation is an event; design is a process

When people say that AI created an image in ten seconds, they are usually describing the rendering event.

They are not necessarily describing the whole project.

The project may have begun much earlier with questions such as:

  • What is the organisation trying to communicate?
  • Who is the audience?
  • What must the viewer understand immediately?
  • Which visual direction belongs to this project?
  • Which details must remain accurate?
  • Which existing references may be used?
  • What should be excluded?
  • Where will the image appear?
  • Will it contain text?
  • How does it relate to the rest of the campaign?
  • Who is responsible for approving it?

After generation, more questions appear:

  • Did the model understand the scene?
  • Are the people distinct and anatomically credible?
  • Does the environment belong to the intended world or brand?
  • Is the product accurate?
  • Does the image preserve enough negative space?
  • Can it be cropped into the required formats?
  • Is the hierarchy readable at thumbnail size?
  • Does anything need to be rebuilt manually?
  • Are there cultural, accessibility, privacy, or reputational concerns?
  • Does the image communicate what the brief required?

The button sits somewhere in the middle of that process.

It does not contain the process.

This matters because organisations may measure the visible moment of generation and mistake it for the total labour involved.
A designer may spend seconds obtaining a first render and hours determining whether it can become a usable asset.

The speed of one stage does not define the speed of the whole practice.

A rendered image is not automatically a resolved concept

Generative models are extremely capable at producing the appearance of resolution.

A result may look coherent because the model has supplied:

  • balanced colour;
  • plausible lighting;
  • surface detail;
  • a familiar advertising composition;
  • expressive faces;
  • cinematic depth;
  • polished visual atmosphere.

But visual polish can hide unresolved thinking.

Imagine that a company asks for:

A futuristic image about collaboration.

The model may generate several employees standing around a glowing table inside a dark glass office.

The image looks professional.
Yet basic questions remain unanswered.

  • What does collaboration mean in this organisation?
  • Are the people solving a technical problem, reviewing research, teaching one another, making a decision, or merely admiring a screen?
  • Why is the office dark?
  • Why is the table glowing?
  • Does the company actually use anything resembling the technology shown?
  • Is the image intended to communicate trust, innovation, teamwork, speed, inclusion, or institutional authority?
  • Does the visual language distinguish this company from any other technology brand?

The result has the appearance of a concept without necessarily containing one.
The generator has filled missing decisions with familiar visual associations.
It has produced an answer to a question the company never fully asked.

A designer’s role is not merely to improve the image.

It is often to recognise that the concept remains unresolved beneath the polish.

Taste is not the whole of design

Design discussions are often reduced to preference.

  • One person likes an image.
  • Another does not.
  • A manager prefers brighter colours.
  • A designer prefers restraint.
  • A client asks for something more exciting.

This can make design appear subjective in the weakest sense: a contest among personal tastes.

Taste matters, but design is not only taste.
A design decision can be evaluated against a purpose.

For example:

  • Is the headline readable?
  • Does the focal point appear where it should?
  • Is the product shown accurately?
  • Does the image work at the intended size?
  • Does the composition support the message?
  • Does the representation suit the audience?
  • Can the asset be used across the required formats?
  • Does it follow established visual continuity?
  • Is the emotional register appropriate?
  • Does it mislead the viewer?

A designer may dislike an image personally and still recognise that it serves the brief.
A manager may love an image aesthetically while the designer recognises that it fails the project.

That is not creative snobbery.

It is the difference between responding to an image and evaluating it as a piece of communication.

Generative systems make this distinction more important because they can produce images that are immediately attractive.

Attractiveness can distract from suitability.

The Visual Register should help users ask not only:

Do I like this?

but:

Does this do what we decided it needed to do?

The first output has become dangerously persuasive

Before generative tools, an early visual draft often looked unfinished.

  • A sketch revealed itself as a sketch.
  • A rough layout looked provisional.
  • A storyboard frame signalled that more work remained.
  • A generated image may arrive with finished-looking light, texture, colour, and rendering.

That visual completeness can create a false sense of project completion.

The first output becomes persuasive because it resembles the final form.

People begin solving the wrong problem:

Which of these polished options do we like?

before solving the earlier one:

Which visual concept serves the communication need?

This can lock a project into an unsuitable direction.

Once stakeholders see a highly rendered image, it becomes harder to discuss structure abstractly. They may become attached to:

  • a particular face;
  • a dramatic background;
  • a colour treatment;
  • an impressive interface;
  • an accidental object;
  • a pose that does not support the message.

The output gains authority because it is visible.
A designer then has to explain why a visually appealing result should be abandoned.

The more polished the draft, the more this can feel like needless resistance.

But rejecting an attractive image can be one of the most important acts in the process.

The image is not successful because it survived rendering.
It is successful when it survives evaluation.

Selection is work

Generative systems often produce several alternatives quickly.
This is described as abundance.

It can also create a new kind of labour.
Someone has to compare the outputs.

Someone has to notice:

  • which face remained consistent;
  • which environment belongs to the project;
  • which pose creates the right relationship;
  • which version contains subtle distortions;
  • which background introduces an unwanted symbol;
  • which composition leaves sufficient text space;
  • which image looks strong only at full size;
  • which image will survive cropping;
  • which variation communicates the intended message rather than merely looking impressive.

Selection is not passive.
It requires criteria.

Without criteria, the person simply chooses the image that produces the strongest immediate reaction.
That may be appropriate for casual experimentation.

It is unreliable for professional work.

The designer’s judgment often appears in what is discarded.

A viewer sees one final image.
They do not see the fifteen seductive but unsuitable versions that were rejected.

The absence of those images is part of the design.

Rejection is not failure

In a button-centred understanding of AI, rejection can look inefficient.

Why generate an image only to discard it?

Because generation is not the same as approval.

A draft may reveal:

  • that the brief was ambiguous;
  • that the Visual Bible lacked an important rule;
  • that the model associates a key word with the wrong visual cliché;
  • that the intended composition cannot carry all requested information;
  • that a product needs better reference material;
  • that the emotional tone was not sufficiently defined;
  • that a requested format changes the scene more than expected.

A rejected output can improve the project if the reason for rejection is understood.

For example:

Rejected because all four people appeared to perform the same gesture.

This reveals a need for more distinct subject blocking.

Rejected because the office appeared like a generic luxury technology company.

This reveals that the organisational environment needs clearer architectural and material definition.

Rejected because the scene looked celebratory rather than analytical.

This reveals a mismatch between visual energy and communication purpose.

Rejected because the image was composed beautifully but could not accommodate the headline.

This reveals that format and layout requirements entered too late.

The useful question is not:

Did the generator fail?

It is:

What decision was missing, misunderstood, or overridden?

That answer may belong in the next brief.

If it should affect future work, it may belong in the Visual Bible.

Revision is not “trying again”

Regeneration is often treated as the main revision method.

  • The user asks for another version.
  • Then another.
  • Then another.

Sometimes this works.
But repeated generation without diagnosis is not the same as revision.

It is repetition under uncertainty.

A deliberate revision identifies what must change and what must remain.

For example:

Keep

  • the four established characters;
  • the Mediterranean pool environment;
  • the playful friendship;
  • the vertical framing;
  • the bright summer light.

Change

  • give each woman a distinct action;
  • move the camera closer to water level;
  • reduce posed smiling;
  • make the movement feel less synchronised;
  • remove the title text.

That is a revision brief.

It preserves successful decisions while correcting specific failures.

A designer does not merely ask for more images.
They control the relationship between continuity and change.

This is precisely the logic at the centre of The Visual Register:

Store what should remain. Ask what is new.

Revision is that law applied after a draft exists.

Prompting is one part of design, not a substitute for it

Prompt writing can involve real skill.

A strong prompt may require:

  • clarity;
  • sequencing;
  • knowledge of model behaviour;
  • visual vocabulary;
  • anticipation of likely failure;
  • reference management;
  • distinction among essential and optional details.

But prompting does not automatically contain every part of design.

A beautifully written prompt can still be based on:

  • a weak concept;
  • an unclear audience;
  • the wrong format;
  • an inaccurate brand assumption;
  • an unsuitable visual metaphor;
  • a misunderstanding of the product.

Similarly, a designer may begin with an inelegant sentence and still make strong visual decisions through references, sketches, blocking, layout, generation, editing, and review.

The prompt matters.
It should not be mistaken for the entire authorship process.

The Visual Register is being designed to produce a readable visual brief rather than worship the final prompt as the central artefact.
The brief can support prompting.

It can also support:

  • a conversation with a designer;
  • a traditional photoshoot;
  • an illustration commission;
  • a motion storyboard;
  • a campaign review;
  • a layout decision.

The underlying design problem exists before the platform is chosen.

A designer often solves the question behind the request

Clients and managers do not always arrive with a resolved problem.

They may ask:

Can you make this more exciting?

Can we have something premium?

Make it feel innovative.

Make it pop.

We need a visual for leadership.

Can you give us some AI concepts?

These requests contain emotional direction but little operational meaning.

The designer may need to uncover:

  • what is currently failing;
  • who the audience is;
  • what action the organisation wants from them;
  • what premium means in this market;
  • whether innovative means speculative, precise, efficient, experimental, or simply modern;
  • what the image must contain;
  • what the image must avoid;
  • how it fits into the wider communication system.

This is design labour.
It often happens through conversation before anything is rendered.

When AI is introduced, this labour can become even less visible.

  • The manager may assume the model will turn the adjective into a concept.
  • The model will certainly produce an interpretation.

That does not make the underlying request complete.

The Visual Register’s required scene seed and purpose fields are partly intended to expose that missing work early.
A request does not become clear merely because the generator is willing to answer it.

“Make it pop” is not a visual brief

Some phrases recur because they describe dissatisfaction without identifying its cause.

Make it pop may mean:

  • increase contrast;
  • strengthen hierarchy;
  • enlarge the subject;
  • simplify the background;
  • introduce one accent colour;
  • add more dynamic movement;
  • improve the headline;
  • create more emotional intensity;
  • make the image less generic;
  • distinguish it from nearby content.

These are different interventions.

  • A designer tries to discover which one is actually needed.
  • A generator may respond by increasing saturation, glow, or spectacle.

That may satisfy the phrase while worsening the communication.

The problem is not that the manager lacks technical vocabulary.
The problem is that the workflow does not help translate a vague reaction into a decision.

A useful system might ask:

What currently feels weak?

  • the main subject is not noticeable;
  • the image feels too flat;
  • the colours lack contrast;
  • the scene feels generic;
  • the emotion is too restrained;
  • the composition is crowded;
  • something else.

This does not turn the manager into a designer.
It makes the dissatisfaction more actionable.

Designers do not object to speed

The conflict is often framed incorrectly.

Designers are sometimes portrayed as resisting AI because they prefer slower methods or fear easier production.

But many designers welcome tools that reduce:

  • repetitive masking;
  • background removal;
  • resizing;
  • rough ideation time;
  • tedious asset variations;
  • basic retouching;
  • manual production steps.

The concern is not speed itself.
The concern is the assumption that speed at one stage removes the need for expertise at every other stage.

A designer may be able to produce more options quickly.

They still need time to:

  • inspect them;
  • compare them;
  • repair them;
  • integrate them;
  • test them;
  • explain them;
  • obtain approval.

Acceleration changes the workflow.
It does not turn judgment into an unnecessary delay.

A fast wrong answer is still wrong.

A large number of fast wrong answers may create more review work than one properly directed attempt.

AI can increase the volume of unfinished work

Before generative tools, the cost of producing a visual option naturally limited how many options were created.
Now a team can generate dozens in a short time.

This can be useful during exploration.
It can also overwhelm the decision process.

More options can produce:

  • stakeholder indecision;
  • inconsistent feedback;
  • attachment to different directions;
  • pressure to combine incompatible elements;
  • longer comparison sessions;
  • confusion about which versions are current;
  • repeated requests for “just one more.”

The bottleneck moves.
The organisation may no longer struggle to produce options.

It may struggle to evaluate them.
This makes clear criteria more important, not less.

The Visual Register’s brief and decision notes can provide those criteria.

The question is no longer:

Which image do we like most?

It becomes:

Which image best satisfies the agreed purpose, continuity, constraints, and output requirements?

Abundance without direction creates noise.

Repair work is design work

Generated images often require correction.

The correction may involve:

  • anatomy;
  • hands;
  • eyes;
  • hair;
  • jewellery;
  • clothing;
  • logos;
  • products;
  • architecture;
  • perspective;
  • reflections;
  • repeated people;
  • unwanted text;
  • missing objects;
  • visual clutter;
  • inconsistent lighting.

Some corrections can be made through editing prompts.

Others require:

  • compositing;
  • painting;
  • masking;
  • reconstruction;
  • typography;
  • vector work;
  • colour correction;
  • layout;
  • manual retouching.

This work is often dismissed as cleanup.
But repair decisions can alter meaning.

  • Removing a person changes the group dynamic.
  • Moving a product changes hierarchy.
  • Rebuilding a face restores identity.
  • Changing the crop alters emotional distance.
  • Replacing the background changes the institution being communicated.

The designer is not simply erasing mistakes.
They are restoring the relationship between the result and the intended brief.

A correct logo does not create a correct brand image

One of the easiest corporate misunderstandings is to treat brand consistency as asset insertion.

  • Use the correct logo.
  • Use the company green.
  • Use the approved font.

These are necessary in many contexts.
They are not sufficient.

A visual may use every official asset and still feel unrelated to the organisation.

Brand identity also appears through:

  • subject matter;
  • environment;
  • material language;
  • posture;
  • representation;
  • tone;
  • visual hierarchy;
  • use of technology;
  • degree of formality;
  • relationship between people and space;
  • the kinds of stories the organisation chooses to depict.

A generic image with a logo placed in the corner remains generic.
The designer’s task is often to interpret how the brand behaves within a new scene.

This is one reason the Company Visual Bible matters.
It can extend brand governance beyond files and colour values into operational visual language.

But even the Bible does not remove design judgment.
It gives judgment a clearer foundation.

Layout begins before generation

An image rarely exists alone.

It may need to live inside:

  • a website header;
  • a blog thumbnail;
  • a presentation slide;
  • a report cover;
  • a social post;
  • a vertical story;
  • a printed poster;
  • an email banner;
  • a product page.

Each format changes what the image must do.

  • A beautiful centred portrait may become unusable when a headline needs to appear beside it.
  • A wide environmental scene may lose its subject when cropped square.
  • A detailed group image may become unreadable as a thumbnail.
  • A dark composition may fail behind white text.
  • A generated image containing its own title may become impossible to localise or revise.

Designers often think beyond the isolated image.
They consider the system in which it will appear.

That may require directing the generation toward:

  • negative space;
  • safe crop areas;
  • simpler background regions;
  • asymmetrical placement;
  • strong silhouette;
  • reduced detail;
  • alternate versions.

These choices should enter before the button is pressed.

Otherwise, the designer may have to extend, rebuild, or abandon the image later.

Accessibility does not arrive automatically

A visually impressive asset may still be difficult to use accessibly.

Potential problems include:

  • low contrast;
  • crowded visual hierarchy;
  • illegible embedded text;
  • information communicated only through colour;
  • decorative complexity around important content;
  • essential subjects disappearing at small size;
  • stereotypes used as visual shorthand;
  • imagery that conflicts with accompanying alt text or message.

The model is not necessarily aware of the full interface, audience, or communication environment.
The designer or reviewer must evaluate the asset within its actual use.

Accessibility is not a final checkbox placed after visual creation.

It can affect:

  • composition;
  • contrast;
  • subject size;
  • information density;
  • text placement;
  • colour choices;
  • whether generation is appropriate at all.

That evaluation is part of design.

Cultural interpretation is part of the work

An image can satisfy the literal brief and still carry unintended cultural meaning.
Clothing, architecture, gestures, body language, colour, symbols, and social roles may be interpreted differently across audiences.

A generator may combine elements that appear visually plausible but are culturally incoherent.

  • It may exaggerate familiar motifs until a place becomes a stereotype.
  • It may represent diversity as a staged arrangement of bodies rather than a believable environment.
  • It may turn religious, regional, or historical references into decoration.

A designer or knowledgeable reviewer may need to recognise:

  • what has been flattened;
  • what has been misplaced;
  • what has been fetishised;
  • what has become generic;
  • what should be removed;
  • what requires consultation.

This cannot be solved by adding the word respectful to a prompt.

Respect requires decisions, context, and review.

The Visual Bible may store cultural boundaries and reasons behind them.

The present image still needs human judgment.

Professional judgment includes knowing when not to generate

Design work is not only the ability to produce an image.
It includes deciding whether an image should be generated at all.

A team may be dealing with:

  • a sensitive event;
  • a real employee story;
  • a community whose representation matters;
  • a product requiring exact technical accuracy;
  • an image that could be misleading if fictionalised;
  • a subject better served by documentary photography;
  • an established illustrator whose authorship is central to the project.

In those cases, generation may be:

  • inappropriate;
  • insufficient;
  • useful only for internal exploration;
  • one component of a larger process.

A button cannot make that judgment.
A designer, editor, manager, or organisation must.

The Visual Register is not intended to route every visual request toward an image model.

Its human-readable brief should remain useful when the best execution method is:

  • photography;
  • commissioned illustration;
  • diagramming;
  • traditional design;
  • stock imagery;
  • no image at all.

A system that treats generation as mandatory has confused a method with the purpose.

The designer’s role may move upstream

As generative tools become more capable, some manual production tasks may decrease.

This does not mean designers disappear.
Their value may become more visible earlier in the process.

They may contribute through:

  • creating Visual Bibles;
  • defining visual systems;
  • establishing environment and character references;
  • designing prompt structures;
  • directing campaign treatments;
  • training teams to make better requests;
  • evaluating outputs;
  • protecting continuity;
  • developing review criteria;
  • deciding which tasks may be decentralised;
  • creating reusable templates;
  • documenting reasons behind visual rules.

This is not merely a defensive redefinition of design.
It reflects where judgment becomes most valuable when production is abundant.

If everyone can generate, the scarce capability becomes knowing:

  • what should be generated;
  • how it should belong;
  • what should be rejected;
  • what needs correction;
  • what should never be delegated.

The Visual Bible itself is design work

A Company Visual Bible does not appear by clicking a button.

Someone has to determine:

  • what the organisation’s visual identity actually is;
  • which parts of the existing brand guide remain relevant;
  • how the workplace should be represented;
  • how products must appear;
  • what visual clichés should be avoided;
  • what degrees of fictionalisation are acceptable;
  • which references are authoritative;
  • which rules are locked;
  • which defaults allow variation;
  • why each rule exists.

That work may involve:

  • designers;
  • brand leads;
  • communications teams;
  • subject experts;
  • legal or safety review;
  • leadership;
  • employees who understand the real workplace.

The Bible is not a magical repository of objective company truth.
It is an artefact of organisational decision-making.

The Visual Register can provide structure.
The company still has to decide what belongs inside it.

The brief is not the final design either

A strong visual brief creates alignment.
It does not determine every pixel in advance.

Design still involves interpretation.

  • Two designers may respond to the same brief differently.
  • Two generators may produce different results.

The brief sets the field:

  • purpose;
  • audience;
  • continuity;
  • message;
  • requirements;
  • exclusions.

Within that field, the creator makes choices.
This is important because a system designed to make work more legible can become over-controlling.

Not every decision needs to be specified before creation.
Some emerge through exploration.

The Visual Register should preserve intentionality without demanding that every creative possibility be resolved through a form.
The brief should provide direction, not suffocation.

The decision record should show judgment, not theatre

The Visual Register may preserve lightweight notes such as:

  • why a draft was rejected;
  • what changed;
  • which rule was updated;
  • who approved the final direction.

This can help managers understand the process.
But the record should not become performative paperwork designed to prove that creative work was sufficiently difficult.

  • The goal is not to maximise the number of recorded decisions.
  • The goal is to preserve the decisions that explain the work.

A useful note might say:

Rejected because the generated team appeared to pose for the viewer rather than collaborate with one another.

That explains a real visual problem.

An unhelpful system might require the designer to document every prompt variation, every crop test, and every discarded adjective.
That would replace invisible labour with administrative labour.

Legibility should support the work.
It should not consume it.

A manager needs enough visual literacy to review responsibly

Managers do not need to become designers.

But if they approve visual work, they need enough literacy to evaluate more than personal preference.

They should be able to ask:

  • What was this image intended to communicate?
  • Which part of the brief does this version satisfy?
  • What makes this direction belong to the brand?
  • Is the subject clear at the intended size?
  • What information may be misunderstood?
  • Why was the previous version rejected?
  • Does the image require specialist correction?
  • Is the image ready for production, or is it still exploratory?

The Visual Register can support these questions by presenting the brief and decision notes alongside the output.

This changes review from:

I like version two.

to:

Version two communicates collaboration more clearly, preserves the product accurately, and leaves enough room for the heading.

Personal response still matters.
It is no longer the only language available.

The requester also performs design decisions

Not every design decision belongs to the designer.

A requester decides—or should decide:

  • the purpose;
  • the audience;
  • the message;
  • required content;
  • practical use;
  • deadline;
  • approval authority.

When these are missing, the designer must invent them or pause the project.
Generative tools can conceal this omission because an image can still be produced.

The requester then sees a polished output and assumes the brief was sufficient.
The Visual Register’s composer should make missing requester decisions visible before generation.

For example:

What is this image for?

What should the audience understand or feel?

What must be accurate?

Where will it appear?

These are not technical design questions.
They are project-definition questions.

Requiring them is not obstructive.
It prevents unresolved management work from being transferred silently into the creative stage.

“Just make a few options” can be expensive

Generating options may be fast.
Reviewing them is not always fast.

Suppose a manager requests ten concepts without a defined purpose.
The team now has ten polished directions.

Each may imply a different:

  • audience;
  • visual identity;
  • workplace;
  • emotional tone;
  • hierarchy;
  • campaign strategy.

Stakeholders select parts from several:

We like the people from three, the room from six, the lighting from nine, and the colour from two.

The designer is then asked to combine incompatible decisions that were never meant to belong together.

More options did not clarify the project.
They multiplied unresolved choices.

A better process may begin with two or three defined directions based on an agreed brief.
The purpose of options is not to avoid deciding.

It is to compare meaningful alternatives.

Design creates relationships among decisions

A prompt may contain many correct ingredients:

  • four people;
  • a laboratory;
  • holographic screens;
  • warm stone;
  • a company logo;
  • dramatic light.

Design concerns how those ingredients relate.

  • Are the people using the screens or merely standing near them?
  • Does the technology support the task or dominate the scene?
  • Does the stone architecture create institutional warmth or visual heaviness?
  • Is the logo integrated plausibly or pasted decoratively?
  • Does the light reveal the work or obscure it?
  • Does one person hold the focal role while others remain legible?

A generator can compose these relationships.
A human still needs to judge whether they are appropriate.

This is why a checklist of included elements cannot fully define success.

An image may include everything and still feel wrong.

The relationships are wrong.
Design lives partly in those relationships.

The strongest image may be the least spectacular

Generative systems reward spectacle.

Large screens, dramatic skies, glowing effects, extreme depth, elaborate architecture, and cinematic poses produce immediate visual impact.

Professional communication often needs restraint.

A strong image may require:

  • fewer elements;
  • clearer hierarchy;
  • more believable interaction;
  • quieter light;
  • greater accuracy;
  • less visual noise;
  • more room for typography;
  • a less glamorous but more credible setting.

The designer may choose the image that looks less impressive in isolation because it works better in context.
This can be difficult to explain when stakeholders equate visual intensity with quality.

The Visual Bible and brief can help establish restraint as an intentional rule rather than a lack of ambition.

For Mithaq Praxis, for example, technology should feel purposeful rather than spectacular.

  • A gigantic floating interface may look futuristic.
  • A smaller, carefully integrated system may communicate the institution more accurately.

The design decision lies in choosing credibility over spectacle.

Design includes continuity across many images

A single generated image may look excellent.

  • A series may still fail.
  • The people change subtly.
  • The environment drifts.
  • The palette moves.
  • The product proportions vary.
  • The mood becomes inconsistent.
  • Every asset appears to belong to a different campaign.

Designers think across the set.

They consider:

  • repetition;
  • variation;
  • rhythm;
  • hierarchy among assets;
  • recognisable lineage;
  • when to repeat a motif;
  • when to introduce contrast;
  • how each image contributes to the whole.

This is where the Visual Bible becomes especially important.

It stores the persistent identity.
But someone still needs to decide how that identity behaves across the series.

Continuity is not produced by copying the same prompt repeatedly.
It is designed through controlled sameness and controlled difference.

The work remains human-led even when much of the image is generated

A human-led process does not require that the human manually create every visible element.
It requires that meaningful authority remain with the person or team.

The human decides:

  • why the image exists;
  • what it communicates;
  • which world or brand it belongs to;
  • what must remain consistent;
  • what may change;
  • which output succeeds;
  • what requires correction;
  • whether generation is appropriate;
  • when the work is finished.

The model contributes:

  • visual interpretation;
  • rendering;
  • variation;
  • synthesis;
  • surface production.

These contributions can be substantial.
Acknowledging them does not erase the human role.

The mistake occurs when the existence of a generated image is used to imply that the human role was unnecessary.
The Visual Register is intended to preserve and expose that role without pretending the machine did nothing.

The button is not the enemy

The title of this article is not an argument against Generate buttons.

Buttons are useful.

A good button can make a complex system accessible.

The problem is not that image generation has become easy to initiate.
The problem is the cultural story built around that ease:

If the image appeared quickly, the work was simple.

That story harms designers.
It also harms organisations.

  • It encourages vague briefs, careless approvals, unrealistic deadlines, and weak visual systems.
  • It discourages people from learning how images communicate.
  • It turns judgment into an inconvenience that delays instant output.

The answer is not to make tools deliberately difficult.
It is to stop hiding the parts of the work that remain necessary.

The Visual Register should make the process legible

The Visual Register will not force every project through a complicated production pipeline.

Its contribution can remain modest.

It can preserve:

The intention

What the human wanted to make.

The continuity

Which Visual Bible and rules applied.

The brief

How the scene, audience, message, format, and constraints were defined.

The revision note

Why an output changed or was rejected.

The status

Whether the work remains a draft, has been selected, or is approved.
That is enough to show that the final image emerged through more than a button press.

The goal is not to create paperwork around creativity.
It is to provide a visible spine for the decisions that already exist.

A practical review language

One useful outcome of the system would be a shared language for reviewing generated work.

Instead of:

I don’t like it.

the reviewer might say:

The emotional tone is too celebratory for the subject.

Instead of:

Make it better.

they might say:

The product is visually subordinate to the environment, but it needs to be the primary focal point.

Instead of:

It doesn’t feel like us.

they might say:

The dark blue holographic styling conflicts with the Company Visual Bible’s bright, credible workplace language.

Instead of:

Can you try again?

they might say:

Preserve the current composition, but replace the posed expressions with task-focused interaction and reduce the decorative interface elements.

This language does not eliminate disagreement.
It makes disagreement more useful.

A practical production distinction

The Visual Register can also help teams distinguish among several stages that are often collapsed together.

Concept

The underlying visual idea.

Brief

The documented purpose, scene, continuity, and requirements.

Draft

An exploratory execution of the brief.

Selected direction

The version chosen for further development.

Production asset

The corrected, formatted, accessible, and approved final material.

A generated image may be a draft.
It may occasionally be close to a production asset.

The two should not be assumed identical.
Making the stage visible can reduce the expectation that every first render is immediately publishable.

The measure of success is not how little the human did

AI products are often marketed by demonstrating the disappearance of labour.

  • One sentence becomes an image.
  • One click becomes a presentation.
  • One request becomes a campaign.

This can be impressive.

It also encourages a strange measure of success:

The less the human contributed, the better the tool.

That is not the measure I want for The Visual Register.
The aim is not to maximise human effort either.

The aim is to preserve the work that should remain human:

  • intention;
  • meaning;
  • judgment;
  • responsibility;
  • continuity;
  • final approval.

And reduce the work that does not need to be repeated:

  • restating established character details;
  • reconstructing the same palette;
  • copying the same avoid list;
  • re-explaining the same environment;
  • manually assembling the same brief structure.

The system succeeds when it makes human decisions clearer and clerical repetition lighter.
Not when it makes the human disappear.

Design is the reason the button’s output becomes useful

A button may generate:

  • an image;
  • a variation;
  • a texture;
  • a composition;
  • a starting point;
  • an unexpected idea.

Design determines whether that material becomes:

  • communication;
  • identity;
  • a campaign;
  • a book cover;
  • a credible corporate asset;
  • a coherent world;
  • a responsible public image.

The generated material may be powerful.
It remains material.

  • Someone still has to decide what it is for.
  • Someone has to determine what survives.
  • Someone has to recognise what is wrong.
  • Someone has to say when the image belongs—and when it does not.

That is the work The Visual Register is being built to support.
Not because designers need a system to prove that their work is difficult.
Because teams need a system that remembers the work exists.

The button produces the image; people produce the standard

A model can generate faster than most people can describe what they want.

That speed is useful.

It also means that organisational standards, creative judgment, and visual literacy must exist before the outputs arrive.
Otherwise, the company becomes surrounded by images it can produce but cannot properly evaluate.

The designer’s contribution is not an obstacle between the organisation and instant creation.

It is part of what prevents instant creation from becoming instant incoherence.

  • The Visual Bible stores the standard.
  • The requester supplies the need.
  • The composer makes the brief legible.
  • The model generates.
  • The designer, creator, or reviewer evaluates what came back.

No single stage contains the whole practice.

And no matter how quickly the image appears, the design was never inside the button.

© 2026 • MITHAQ PRAXIS • CC BY-NC-ND 4.0 Unless Otherwise Stated.