Case study · Type design · IDC School of Design, IIT Bombay

Interlocking देवनागरी

A display typeface where letters make room for each other — and the two-semester search for the rules that let them. My M.Des thesis, guided by Prof. Girish Dalvi.

◆ Published — Typography Day 2021 ◆Featured on D'source

RoleResearch, type design, writing — solo

Timeline2020–21, two cycles

GuideProf. Girish Dalvi

ContextM.Des Communication Design

OutcomeA display glyph set + documented system

The itch: every other script had a way in

Latin has a whole genre of interlocking display type. Arabic calligraphy has been doing it for eight centuries. Devanagari — one of the most-used scripts on earth — had beautiful one-offs, and no written rules.

The references were everywhere once I started looking. Shiva Nallaperumal's Calcula generates over 6,000 interlocking sets by manipulating letters around just five "intruder" glyphs. Square Kufic has filled mosque walls since the 12th century, black and white balanced so tightly that no gap exceeds one square. Diwani Jali interlaces a whole sentence into one-piece artwork.

And in Devanagari? S.M. Wadisherla's lettering interlocked, maybe without meaning to. Ranjana script's Kutakshar monograms stack letters by defined rule. Sign painters on Indian streets had been overlapping wide letterforms for decades. The instinct existed everywhere. The system didn't.

"In interlocking fonts, the glyphs allow the font to visually manifest as ornamentation, pattern, or a means to save space — achieved by glyphs being overlapped, fitted like a jigsaw, or following a grid."

The working definition I distilled from the research — the brief I gave myself

Fifteen ways to interlock a script

Before choosing an answer, I wanted the whole question. So I drew — one word at a time, each word a different theory of how Devanagari letters could share space.

Plump overlapping forms. Vertical stacking borrowed from old conjuncts. A jigsaw of diagonals. Kufic mazes. Ornamented iron gates. Letters written sideways for narrow streets. Each sketch got its own system-of-interlocking, because a pretty word without rules is lettering, not type.

Sketch of maujooda in plump overlapping letterforms — **मौजूदा** — plump characters, broken sirorekha; everything can overlap, but it reads like tight tracking. Rejected for low novelty.

Taajdaar-e-haram in monolinear rounded letterforms — **ताजदार-ए-हरम** — big round loops create white space for neighbours to move into. Comic, naive, charming.

Blue calligraphic words drawn with a canted nib — **टनटनाट · आक्रोश · गीतांजलि** — a canted-nib font in four rows; weight shifts up or down by neighbour. Space-efficient.

A Sanskrit sentence in mint interlocked letterforms on black — **अस्माकं कार्याणि…** — the canted-nib idea stress-tested on a full Sanskrit sentence.

Mushkil samjhauta in thin cyan monoline forms — **मुश्किल समझौता** — elegance through omission. Beautiful, and nearly impossible to systematise.

Jhakaas in crooked chunky outline forms — **झकास** — crooked, chunky, fluid. When every letter is a little deformed, deformation becomes the style.

Darwaza as a red Square-Kufic maze — **दरवाज़ा** — Square Kufic, translated. Letters reduced to basics, interlocking on a strict grid.

Vertical Devanagari words written top to bottom — **Vertical Devanagari** — for narrow streets where signboards run top-to-bottom. A real problem, but essentially a monospace brief. Cut, reluctantly.

Feedback came from my guide, colleagues, and Noopur Datye of Ektype. The advice that stuck: execution defines how good a style is — the choice of style is a small part of it. Eliminate what can't work, then pick what you want to live with. One sketch kept pulling.

Pencil sketch of dravid utkal bang in diagonal jigsaw letterforms — The oneद्रविड़ उत्कल बंग — glyphs shaped so each leaves a space the next can occupy. Diagonals, triangles, a broken sirorekha. A jigsaw you could read.

The first font — and the honest jury

Project II turned the jigsaw sketch into a working display face. It also taught me the most useful lesson of the whole thesis — by failing in a very specific way.

The letterforms found their skeleton on Otl Aicher's 1972 Olympic pictogram grid — the diagonals and triangles fell into it almost on their own. From there: modulated strokes, Art Nouveau-ish languid joins, stylised tails, heavy wedges to balance the whitespace, chiselled kana.

The word interlocking on its construction grid — PlateThe word इंटरलॉकिंग on its own construction grid.

Annotated anatomy of the typeface — PlateThe anatomy, annotated — angular turns, chiselled kana, heavy wedges, filled knots, languid neck joineries. Every feature creates or fills space.

The word aadarsh set twice with different interlocking — Plateआदर्श, twice — same word, two tightnesses. How far should the lock go?

Tamaso ma jyotirgamaya in the typeface — Plateतमसो मा ज्योतिर्गमय — the first warning: with few interlocking pairs, the rhythm went monotonous.

Dakait in lavender jigsaw letterforms on purple — Plateडकैत — the concept at its best: three letters locked like a badge.

Uday reduced to jigsaw pieces plus two rendered versions — PlateThe logic, reduced: उदय as three puzzle pieces. Each glyph modifies itself just enough — visually similar, functionally different.

Glyph set beside the grey jigsaw shapes — PlateThe glyph set, each letter beside the puzzle-piece it secretly is.

Dakait poster with crossed rifles — PlateIn use — a डकैत title card. Stone engraving, saga posters, rock albums: that was the register.

Then the jury spoke.

The interaction between glyphs is so subtle that the surprise an interlocking font promises never arrives.
The font is half-baked — where the system fails, there are no alternate glyphs to catch it.
The visual feel drifted during development; the charm of the lettering got lost on the way to the font.
Some glyphs look odd because the grid was obeyed when it should have been broken.

They were right on every count. I had a font. I no longer had the idea. So for Project III I made the call that felt expensive and was actually cheap: keep the question, discard the answer, start again.

Starting over, on purpose

Round two began with three corrections written on the wall: make the interlocking direct — forms of a letter may differ, as long as they're consistent. Balance every aspect at once — visual grammar, system, conjuncts, grey value. And go where the vertical space is.

A condensed skeleton gives Devanagari room to move: tall narrow glyphs, small ascenders and descenders, vertical space to spend on stacking and locking. New explorations tested that thesis.

Mukhalif in white condensed flat-pen forms on black — Plateमुखालिफ़ — condensed, flat-cut pen. Light horizontals keep vertical interlocking clean. This one became the font.

Ek aur ek gyarah in condensed monoline on mauve — Plateएक और एक ग्यारह — the condensed thesis, monolinear version.

Aashanka in heavy reversed-contrast forms — Plateआशंका — a reversed-contrast detour. Novel, but it locked poorly. Filed away.

The most useful advice of this cycle: draw the basic glyph set first, define the system of interlocking second, then merge the two. One phrase tracked the whole journey from sketch to font:

Concept sketch of akl par patthar padna — Beforeअक्ल पर पत्थर पड़ना — the concept sketch.

The same phrase in the developed typeface — AfterThe same phrase in the developed font — one interlock traded for a vertical stack, because the lock was creating the very whitespace it existed to remove.

The system: rules before beauty

The core idea of the final font is that a glyph is not a shape. It's a set of states — compressed, extended, half-height — each optically corrected to look like one calm letter.

Interlocking then becomes arithmetic: n units of space created by one glyph must be filled by n units of another. A lock that leaves whitespace behind has failed at its only job.

Five structural forms of the letter a — Plateअ in five structural forms — parts compressed or extended, then corrected to stand alone.

Ten possible forms of aa, one discarded — PlateAdd ा at two heights and आ has ten forms. One gets discarded — it makes space nothing else can fill.

Grid of forty usable states of the letter ka — Plate — the countक, exhaustively: eight base forms × matras = forty-eight candidates, eight discarded, **forty usable states** — the most any letter carries. The tinted ones are the rejects.

Four kinds of interlocking got names and rules. Watch one sentence — बिल्ली और दूध की रखवाली, "the cat guarding the milk" — go through all four:

1DefaultNo locks. The control — proof the font holds up when nothing interlocks.

2Letter with letterऔ opens a space the width of र — so र moves in. n made, n filled.

3Letter with its own matraThe ी folds back into its own consonant. Space-saving inside one character.

The same sentence with matras interlocked

4Vertical stackingSimilar widths stack. The sirorekha breaks on purpose — so सफ़र is never misread as सरफ़.

The same sentence with letters stacked vertically

Accha rasta set two ways showing conjunct treatment — PlateConjuncts behave like any other letter — the rules address parts of letters, so च्छ is just one more shape with edges.

Vertically stacked words with broken sirorekha — PlateThe stacking legibility rule at work: break the headline, keep the reading order.

Underneath the system, quieter craft. Straight condensed stems turned oppressive in stem-heavy words — so every vertical got a barely-there curve, drawn as a tangent rather than a rounded rectangle. Its direction doubles as a legibility cue when letters lock. And a proportion grid holds it all: matra : body : descender at 0.25 : 1 : 0.1.

Before and after of ka and pa showing the tangent curve — DetailThe hand facet — before and after. Red lines mark the drift each stem is allowed.

Proportion grid over billi aur doodh ki rakhwali — DetailThe proportion facet — vertical matra 0.25 : 1 : 0.1, internal verticals at 1 : 30.

And a fourth category that breaks its own rules on purpose: contextual wordmarks — phrases so common they get coded as single words, free to interlock however they please.

Plateराम राम — the everyday greeting as a wordmark, read top-left to bottom-right.

One honest asterisk, kept honest in the report itself: with multiple states per letter, many words have more than one valid setting. I tried to derive rules for a single "right" interlock from first-time readers; the attempts were inconclusive. Verdict as shipped: it's the user's choice — pending a formal legibility study.

The word lagta set four ways — Plateलगता, four ways — default, economical, stylistic, stacked. All legal.

The typeface

A condensed, humanist Devanagari display face. Medium contrast from a flat-cut pen. Open counters, squarish turns, dragged end-knots. Space saved without ever looking squeezed — elegant and casual were the two adjectives it had to earn.

The basic glyph set of the final typeface — PlateThe basic glyph set.

Aadarsh Malviya set in the final typeface — Plateआदर्श मालवीया — the designer's name, set in the design. Note र्श folding into place, and म्ल stacking.

A Hindi couplet set plainly in the typeface

One poem, two settings —

InteractiveNida Fazli's जिधर जाते हैं सब — first as plain condensed type, then with the system switched on. Same poem, same face; the texture is the feature.

Where it honestly ends

This is a thesis, not a retail release — and the report says so out loud. It was also selected as a paper at Typography Day 2021 and featured as a case study on D'source.

Still to do, documented as further work: glyph coverage beyond Hindi and Marathi; akhand conjuncts and dense clusters; kerning and texture under a formal protocol; and the study this font genuinely needs — measuring legibility and reading speed when readers meet unconventional reading patterns, so the rules can graduate from designed to proven.

What the project settled for me is bigger than the font: define the system first, and the beauty has somewhere to live. It's the same instinct I later brought to design tokens at a startup and a design system at Flipkart — just with pixels instead of pencils.

WatchTwo short films from the project — the conference talk and the case-study film.

The paper, presented at the IDC-hosted conference.

"When you have exhausted all possibilities, remember this: you haven't."Thomas Edison — the line the thesis closes on

Credits & record

ProjectInterlocking Devanagari Display Font
Projects II & III, M.Des Communication Design

DesignerAdarsh Malviya — research, drawing, systems, writing

GuideProf. Girish Dalvi, IDC School of Design, IIT Bombay

PublishedTypography Day 2021 · paper & talk

FeaturedD'source case study

Feedback circleNoopur Datye (Ektype) · Athang Samant · Ameya Nikose

Next case studies — Flipkart's design system and Creso, 0 → 1 — are in the works.← Back to the portfolio