Brushless Illustrations

Brushless
All work

Case studyGreek EdTech brand

Ten times faster per illustration, and a roadmap that stopped waiting.

A children's learning app needed thirty to fifty unique illustrations for every course and had one freelance illustrator to make them. Finished courses sat unlaunched behind the queue. We trained on the app's own library: illustrations went from two-to-four hours each to fifteen, and the twenty-eight-letter Arabic alphabet set — quoted at six to eight weeks — was delivered in five days.

Sector
EdTech, Greece
Scope
12 planned courses · 30–50 illustrations each
Turnaround
5 days for the 28-letter Arabic set
Previously
$6.75K–$11.25K per course, 2–4 hrs each
10×faster production per illustration
70%
reduction in illustration cost per course
5 days
for 28 Arabic letters, down from 6–8 weeks
4×
more courses launched in the same window

All figures are actual project outcomes. Client names are withheld under confidentiality agreements.

The situation

One illustrator, and a product roadmap behind them.

An edtech startup in Greece building a mobile learning app for children aged four to ten — language, maths and general knowledge, each course a unique set of illustrated screens and interactive elements.

The app's identity was the illustration: flat vector, friendly characters, clean lines, bold colour, playful without stopping being educational. Every new course needed thirty to fifty original pieces — character poses, icons, scene backgrounds, interactive elements — and the company was expanding into Arabic, which required a new alphabet set where each letter becomes a child-friendly illustrated object.

One freelance illustrator produced all of it. They were good and they understood the style completely; the constraint was arithmetic. At two to four hours an illustration, thirty to fifty per course, the app could publish courses at exactly the rate one person could draw them.

A single-illustrator ceiling
The whole visual output depended on one person working one course at a time, with no way to add capacity without adding style drift.
$80K–$135K a year
Twelve planned courses at $6,750–$11,250 each in illustration fees alone.
Courses written and idle
Content that was finished and signed off, waiting on illustrations before it could ship.

What changed

Per illustration, per course, per project.

The same three measures the team was already tracking, before and after. The Arabic set is listed on its own because it was the project that had been quoted and shelved.

Per-illustration time
2–4 hoursone freelancer, one illustration at a time
15–25 minincluding creative review and refinement
Per-course cost
$6.75K–$11.25Kfreelance illustration fees for one course
$2K–$3.4Ka 70% reduction across the twelve planned courses
The Arabic alphabet set
6–8 weeksfreelancer estimate for 28 illustrated letters, at ~$15,000
5 daysall 28 letters delivered, with the variations

Consistency came out higher than the single-illustrator baseline rather than lower — subtle drift is what happens to one person's hand over many months, and a model trained on the library does not have months.

What we built

Trained on their library, not on a description of it.

We started from the illustrations the team had already made. Line weights, palette, character proportions, shading, the compositional patterns that repeat across screens — the model learned all of it from the assets themselves rather than from a written brief about them, which is why the outputs sit inside the style instead of near it.

The Arabic alphabet was the harder half and it was not a styling problem. Each letter had to be legible as an Arabic character AND readable as a playful illustrated object, and the two constraints pull against each other. We worked through the pedagogical requirements with the company's educational content team before training for it, so the model was producing against the teaching standard and the visual one at the same time.

Their designers stayed in the loop the whole way: generate, select, adjust, refine. The volume moved; the editorial control did not.

How it ran

  1. Study the existing library

    Every illustration the app had already shipped, read for line weight, palette, proportion, shading and composition.

  2. Define the teaching constraint

    For the Arabic set, the pedagogical requirements written down with the content team: what makes a letter recognisable as a letter while it is also an object.

  3. Generate against both

    Illustrations produced to the visual style and the educational clarity standard at once, rather than styled first and checked afterwards.

  4. Review and refine

    The design team selected and adjusted; the content team verified each screen was educationally sound before it shipped.

“Our product roadmap was being dictated by illustration capacity, not by content readiness.”
Head of Product, the startup

Since

The same model now produces the promotional material, the app store screenshots and the social content, with no additional setup — the style was trained once and the app's marketing happens to be in it. Four courses launched in the window that used to illustrate one.

Notes
  1. Client not named under a confidentiality agreement. The sector, the audience and the shape of the engagement are described because they are what the work was.
  2. No illustration from this engagement is shown, and none will be. Every output is the client's own brand property in the client's own style — there is no redaction that leaves an illustration publishable the way removing a lockup leaves an ad board publishable. This page is set without visuals for that reason rather than for want of files.
  3. Figures are the client's: the two-to-four-hour and six-to-eight-week estimates are the freelancer's own quotes, and the per-course range covers the twelve courses planned at the time of the engagement.

Need more illustrations than your team can draw?

We can train on the style you already have and build a system that produces on-brand illustration at whatever volume the roadmap actually needs.