2.5 The Persona Brand Kit: Visual Identity

An account becomes recognizable through repetition of a small number of visual decisions. Most operators make hundreds of them per week, differently every time, and then wonder why their feed does not look like one account.

The brand kit is the short list of decisions you make once. It is not a mood board. It is closer to a set of constraints, and its main benefit is economic: constraints make production cheaper. Every choice you fix is a choice you do not remake on every post.

The kit has two layers

The persona's style — what she wears, how she presents herself. This belongs in the character bible (2.3), because it is a fact about her.

The production style — how she is photographed, colored, and framed. This is the brand kit, and it is a fact about your pipeline.

Keeping them separate matters. Changing the production style is a visual refresh; changing the persona's style is a character decision with continuity consequences.

Palette

Three dominant colors and two accents. Not ten. A palette of ten is a mood, and a mood produces an inconsistent grid.

Practical constraints on choosing them:

One dominant neutral — the color most of the frame is. This is usually the environment and the grade, not the wardrobe.
One warm and one cool, so seasonal variation has somewhere to go.
Two accents used sparingly, in props, nails, a bag, a caption card. Accents are what make a grid recognizable without being obvious.
Check the grid, not the swatch. Lay out nine past posts. If the grid reads as one palette, it works. Swatches lie; thumbnails do not.

The tone curve is part of the palette. More than any single color, a consistent grade is what makes a set of images read as one account. Decide the curve — contrast, warmth, black point, grain — and apply it as a preset in the post-processing stage from 5.5 rather than by eye each time.

The single highest-return item in this kit is a saved preset. A consistent grade applied to every image does more for recognizability than any wardrobe decision, and it costs nothing per post. If you do only one thing in this chapter, do that one.

Typography, with a warning

If the persona publishes graphics with text — quote cards, list formats, story text — fix two fonts and a size hierarchy.

But note the constraint that governs this whole area: generated text inside images is unreliable. Text rendered by an image model is frequently malformed, and it is one of the tells from 1.4. The practical rule:

Text over an image is composed after generation, never by the model. You place it in the edit step.
Captions and story text belong to the platform, not the image. Fewer rendering problems and better accessibility.
If an image genuinely needs in-frame text — a sign, a menu, a screen — treat it as a QA , not as a lucky result. Chapter 5.6 makes it a mandatory check.

Visual signature

One recurring element the audience can identify without reading the caption. Examples of the kind of element, not recommendations:

A recurring framing habit (always slightly low, always a bit off-center)
A signature prop that appears in many posts
A specific treatment: film grain, a light leak, a shallow depth of field
A recurring location that functions as a set
A compositional rule (the subject always on one third, a lot of negative space)

One element. Two is a style; five is a mess. The signature works because it is the thing that appears when everything else varies.

The test: show nine thumbnails to someone and ask which account they belong to. If they can tell, the signature is doing its job.

Framing rules

Fix the answers to four questions and write them down, because they are the choices you would otherwise remake hundreds of times:

Shot sizes. Two or three that you use. A medium shot and a close-up is a complete system; adding ten more is how accounts lose coherence.

Camera height. Eye level, slightly below, slightly above. Pick one as dominant. Height changes the character's perceived status more than any other variable.

Focal length look. Wide and environmental, or long and compressed. This affects the face (Part IV) as much as the composition — a wide lens close to the face distorts features, and that distortion must be consistent or the identity will read differently across posts.

Aspect ratios per platform. Decide once, per platform. Chapter 8.1's platform map sets which.

Lighting rules

Three conditions you actually use, each with a defined character. From 4.6's battery:

A baseline — window daylight or a softbox. Most posts.
A flat frontal — ring or on-camera flash. Unforgiving, useful for texture checks.
A hard directional — editorial flash or late sun. Sparingly, for contrast.

The rule that matters: do not mix conditions within a single post or carousel. A set where light direction changes between frames reads as compiled rather than taken, and it is one of the environmental tells from 1.4.

Record the setup used for the best-performing posts. That is not sentiment; it is evidence about what the audience responds to, and it belongs in the reference board from 1.6.

Where each element is enforced

A kit that lives only in a document is a preference. Each element needs an owner in the pipeline:

Kit element	Enforced at	Chapter
Palette and grade	Post-processing, as the saved preset	5.5
Framing and shot sizes	Prompt templates and batch configuration	5.4
Lighting conditions	Scene specification, per batch	5.2
Text placement	Edit step, never generation	5.5
Signature element	The publishing checklist	5.6
Aspect ratios	Export profile, per platform	8.1
Face identity	Not here — Part IV	4.5, 4.6

That last row is deliberate. Face identity is not a brand-kit decision; it is the training and QA program in Part IV. Confusing the two produces a persona whose look is consistent except for the face, which is the worst possible distribution of consistency.

Why this makes production cheaper

Four concrete savings.

Decisions collapse. Instead of choosing framing, grade, and palette per post, you choose from three pre-made combinations.

Batches become possible. Chapter 5.4's pipeline requires consistency by definition. A kit is what makes a batch look intentional rather than repetitive.

Quality control becomes checkable. "Is the grade the saved preset?" is a yes/no question. "Does it look right?" is not, and it produces a different answer every day.

Refreshes are deliberate. When the account needs visual evolution, you change one kit element and date it. Without a kit, refresh happens accidentally and continuously, which reads as drift.

Where this goes wrong

A palette of ten colors. It is a mood, not an identity.

Choosing colors from swatches. Judge from a nine-image grid.

No saved preset. You will grade by eye, differently each time, and the feed will drift.

Text generated inside the image. The most common avoidable tell in the medium.

Five signature elements. Recognition requires one thing to be constant.

Mixing lighting within a carousel. It reads as assembled from different sources.

Treating the kit as decoration. It is a production constraint system with a cost payoff.

The kit in one page

Consolidated, so it fits on a wall next to your monitor:

Palette: three dominant, two accents, judged on a nine-image grid
Grade: one saved preset, applied to everything
Text: composed in the edit step
Signature: one recurring element
Framing: two or three shot sizes, one dominant camera height, one focal look
Light: three conditions, one baseline, never mixed within a post
Exports: fixed ratios per platform

Seven lines. When the kit grows past a page, it has stopped being a constraint system and started being a document, and the constraints will stop being followed.

The gate

Three dominant colors and two accents — and does a nine-image grid prove it?
Is the grade saved as a reusable preset?
What is the single visual signature?
Which shot sizes, camera height, and focal look are fixed?
Which three lighting setups are in rotation, and which is the baseline?

What you leave with

A two-layer kit that separates the character's style from the pipeline's, a palette chosen against the grid rather than the swatch, the saved preset as the highest-return single item, a text rule that keeps rendering tells out of your feed, one visual signature, fixed framing and lighting rules, and four concrete ways the kit makes each post cheaper to produce.