The hand-drawn aesthetic in 2026

Cora's Atlas could have been a sleek vector game. The art could have been a palette of three brand colours, a serif headline, a glass panel with a six-pixel blur. We had the templates open in a tab for about an hour. Then we closed the tab. What we shipped instead is hand-drawn portraits, soft watercolour maps, parchment cards with margin marginalia in walnut ink. This is the dispatch about why.

1. The default we rejected

The SaaS default of 2026 is easy to describe because every landing page is doing it. Glass morphism over a dark gradient. A hero image that is either a synthetic 3D render or a stock photograph of a hand holding a phone. A neon-mint accent colour. A serif sub-headline trying to add warmth to a stack of utility classes. Vector minimalism, sharp shadows, tech-bro confidence in two columns. We are not going to name a product. There is no point. Every product looks like every other product, and that is the diagnosis, not a complaint.

We are not above the default. We started there. Our first three landing-page comps had a deep navy gradient, a glass card, a thin sans-serif headline, and an accent colour that was, embarrassingly, neon mint. They looked competent. They also looked like everyone. A puzzle game that promises a quiet afternoon on a parchment page cannot show up at the door wearing the uniform of the platform that wants you to come back tomorrow. The default was the wrong dialect.

2. The choice

So we leaned into warmth. Hand-drawn portraits for the main cast. Soft watercolour washes for the regional maps. Parchment cards for the carnet pages and the puzzle frames. Walnut-ink line work in the margins for the marginalia, the lists, the small numbered annotations Cora makes when she is bored at a long table.

Hand-drawn does not mean amateur. It means every line was a decision. A vector mark is a setting in a panel; a drawn mark is a choice somebody made on a Tuesday afternoon, holding a brush, with the wrong music on. The two marks read differently because they are different. A reader who could not articulate why often feels it anyway. They slow down on the page.

3. The honest part: AI in the concept loop, human in the redraw

The studio policy is to be honest about AI use where it is relevant, so here is the relevant version. The concept passes for some assets — early portrait studies, regional palette explorations, draft watercolour washes — were generated with image-model tools and reviewed against the canon brief. The concepts were never the shipped art. Each one went into a redraw pass by a human, sometimes the same person, sometimes a different one, where the line work was rebuilt, the proportions corrected against the character bible, the period detail adjusted (the wrong button shape, the wrong shoulder line, the anachronism the model could not know about), and the tone re-keyed to the parchment palette.

We are not claiming a hundred percent human craft. We are claiming a loop: model proposes, human disposes. The model is a fast sketchbook. The shipped frame is a human decision. We are saying this out loud because we think in 2026 the studios that lie about it will be found out, and because the loop is the actual interesting thing — neither pure machine output nor a romantic claim of all-hand-made.

4. The tradeoff

The hand-drawn route is slower to ship. A vector portrait can be re-coloured in an afternoon. A hand-drawn portrait, when we need a new variant, takes a small project. Refreshes are not trivial. We cannot promise a calendar; we will not.

The compensation is half-life. A CSS gradient dates in eighteen months. The accent colour that looked sharp last spring looks like a screenshot of last spring this spring. A hand-drawn portrait dates more slowly than that. The line work has a year of half-life, not a quarter. A watercolour wash with three pigments that were already old when the brush touched them ages more like a book than like a deck. We bought slower ageing with slower shipping. We think the trade is correct for the genre we are working in. A puzzle game whose contract is the world holds while you step away from it cannot dress itself in the visual idiom of come back tomorrow.

5. A worked example: the protagonist's portrait, across versions

The portrait of the protagonist at the centre of the saga went through more passes than any other asset. The earliest version was too youthful — early-twenties, smiling slightly, not the right read for a cartographer with seven years away from home. The next version over-corrected — severe jawline, hooded eyes, almost gothic, the kind of face that signals the wrong genre. The third version was too neutral, which is the worst category of all in portrait work: a face that says nothing.

We settled at twenty-eight, carnet in the foreground, gaze a little off-axis, the line work between mélancolique and alert. The shoulders are angled toward something off-frame, which is the read we wanted — a person who is paying attention, not posing. The parchment behind her is the same palette as the carnet pages in the game. The portrait is a frame from the world, not a glossy cover plate sitting on top of it. We do not name the artist or artists in the studio; the studio is anonymous. But the portrait belongs to the people who did the redraws, not to the model that proposed the first sketch.

6. A portrait that holds five weathers

A still portrait would still have been wrong. The protagonist appears across many moments — quiet ones, sharp ones, the morning after a long page in the library, the second before she answers a question she has been avoiding. A single fixed expression would have flattened her across the whole saga.

The portrait is built as a small set of variants — five emotional states the parchment can hold — that the page picks from according to the moment of the story. The shift is subtle. A line in the jaw. A different placement of the gaze. A slightly warmer or cooler wash on the parchment behind her. A reader who is not looking for it may not notice. A reader who is paying attention catches it on the second read and feels the dispatch is, very quietly, looking back.

The five-variant system is more work than a single portrait would have been. It is less work than five separate portraits. The middle path is the medium we are interested in: hand-drawn enough to mean something, systematised enough to ship.

What this is, and what it is not

This dispatch is not an argument that vector design is bad. It is not an argument that everyone should hand-draw. Plenty of products are correctly served by the platform default, and the studios that work in that idiom are doing competent work inside the constraint they chose. Our constraint is different. We are making a puzzle game with a parchment palette, walnut tones, margin marginalia, and a protagonist who carries a brown leather carnet of sixty vellum pages and reads it slowly. The medium has to match the story, or the story arrives dressed for the wrong room.

We picked the slow medium in the year of the gradient. We expect to be slower to ship than the competition that picked the fast medium. We are betting that the visual half-life is worth it, that a reader who lands on a parchment card will breathe one beat differently than the reader who lands on a glass panel, and that the breath is the whole product.