Good bilingual typography in Hindi and English matches the scripts by their visible proportions, not their point size. Devanagari hangs from a headline (shirorekha) that, in harmonised Google Fonts families, sits close to the Latin cap height, and its vowel signs reach well above and below that, so at the same size Hindi looks larger or smaller depending on the pairing and always needs more line height. Pick a family designed for both scripts, check the ratio of the Devanagari headline to the Latin x-height (1.16 to 1.35 in the nine families I measured), size each script to match, and give Hindi lines about 1.6 times the font size.
- Match scripts by visible proportion, not point size.
- In 9 Google Fonts families, the Devanagari headline is 1.16–1.35× the Latin x-height.
- Mukta, Baloo 2 and Rozha One lock the shirorekha to cap height (ratio 1.00).
- Matras span 1.06–1.30 em, so Hindi body text needs a line height of about 1.6.
- Build the baseline grid on Devanagari, the taller script.
- Rule 11 of the Official Languages Rules, 1976 requires both scripts but sets no sizes.
Indian design work is bilingual by default. Central government offices must issue name-plates, sign-boards, letterheads and forms in Hindi and English under the Official Languages Rules, 1976, and most consumer brands add a regional script for reach.
Yet most designers set Latin well and treat Devanagari as an afterthought. The result is the familiar sign where the Hindi line looks either shrunken or shouting, with matras colliding with the line above.
Why does Devanagari look a different size at the same point size?
Point size measures the font's em box, not the letters. Each script's designer decides how much of that box the letters use.
The two scripts have different jobs to fit into the same box:
- Devanagari has to fit a tall headline body plus vowel signs above it and below the baseline;
- Latin only has to fit lowercase, capitals and descenders.
What the eye reads as size
The useful comparison is between the parts the eye reads as "size". In Latin text that is the x-height. In Devanagari it is the body from the baseline up to the headline bar, the shirorekha.
I measured these directly from the font files of nine families on Google Fonts, in units per 1,000 em:
See the data as a table
| Latin x-height (units/1000 em) | Latin cap height (units/1000 em) | Devanagari headline (top of क) (units/1000 em) | |
|---|---|---|---|
| Mukta | 470 | 630 | 630 |
| Hind | 508 | 674 | 642 |
| Baloo 2 | 469 | 613 | 610 |
| Tiro Devanagari Hindi | 477 | 701 | 620 |
| Noto Sans (+ Devanagari) | 536 | 714 | 622 |
| Poppins | 548 | 697 | 740 |
| Yatra One | 478 | 666 | 619 |
| Rozha One | 460 | 560 | 560 |
| Kalam | 511 | 720 | 645 |
Source: Measured by me from the Regular font files in the google/fonts repository (Nov 2026) · Glyph bounds of x, H and the Devanagari letter क. Noto pairs Noto Sans (Latin) with Noto Sans Devanagari; the others carry both scripts in one file.
Three patterns jump out:
- Locked to cap height. In Mukta, Baloo 2 and Rozha One, the designers aligned the Devanagari headline exactly with the Latin cap height (ratio 1.00).
- Set lower. In Noto, Tiro and Kalam, the headline sits lower, at 0.87-0.90 of cap height.
- The outlier. Poppins' Devanagari headline (740) is taller than its own Latin capitals (697), so Hindi set in Poppins next to its English looks noticeably larger.
What does the measured data say about size matching?
If your Latin and Devanagari come from the same harmonised family, set them at the same size and they will match. If you mix families, scale the Devanagari so its headline-to-Latin-x-height ratio stays near the 1.25-1.35 range that most designed pairs use.
| Family (designers, foundry) | Headline ÷ Latin x-height | Headline ÷ Latin cap height | Vertical extent of matras (ि top to ु bottom) |
|---|---|---|---|
| Mukta (Girish Dalvi, Yashodeep Gholap; Ek Type) | 1.34 | 1.00 | 1.12 em |
| Hind (Manushi Parikh, Satya Rajpurohit; Indian Type Foundry) | 1.26 | 0.95 | 1.16 em |
| Baloo 2 (Sarang Kulkarni; Ek Type) | 1.30 | 1.00 | 1.16 em |
| Tiro Devanagari Hindi (John Hudson, Fiona Ross; Tiro Typeworks) | 1.30 | 0.88 | 1.16 em |
| Noto Sans + Noto Sans Devanagari (Monotype) | 1.16 | 0.87 | 1.16 em |
| Poppins (Ninad Kale, Jonny Pinhorn; Indian Type Foundry) | 1.35 | 1.06 | 1.30 em |
| Yatra One (Catherine Leigh Schmidt) | 1.29 | 0.93 | 1.19 em |
| Rozha One (Tim Donaldson; Indian Type Foundry) | 1.22 | 1.00 | 1.06 em |
| Kalam (Lipi Raval, Jonny Pinhorn; Indian Type Foundry) | 1.26 | 0.90 | 1.20 em |
The last column is the one that breaks layouts. From the top of the i-matra (ि) to the bottom of the u-matra (ु), Devanagari needs 1.06 to 1.30 em of vertical space for a single line.
Set it at a line height of 1.2, a common default for Latin headings, and the matras of one line will touch or overlap the next. That is why Devanagari body text usually wants a line height around 1.6, and headings at least 1.3.
In 2012, Harvard University Press commissioned the type designers Fiona Ross and John Hudson, through Tiro Typeworks, to make typefaces for a new book series, the Murty Classical Library of India. Funded by a $5.2 million gift from Rohan Murty, the series prints each classic in a dual-language format, the original script on one page and the English translation facing it. So the fonts had to look traditional, read cleanly for long stretches, and sit beside English without either side looking out of scale. The designers studied 18th- and 19th-century manuscripts and early Indian printing, and met repeatedly with HUP's editors and book designers, at the British Library and at the University of Reading. Publishing began in January 2015; one of the first five volumes was Sur's Ocean, in Hindi. In 2019 Google Fonts asked Tiro to extend the Murty fonts into an open collection, the source of today's Tiro Devanagari Hindi in the table above.
Google Design: The modern Tiro Indic collectionMatch Devanagari size to your Latin text
Which font families pair Hindi and English well?
Use a family designed for both scripts together, because the designers already solved stroke weight, contrast and vertical alignment. Mixing a Latin family from one source with a Devanagari from another almost always needs manual size and weight correction.
- Mukta (Ek Type): a clean humanist sans in seven weights, from ExtraLight to ExtraBold, and the headline locked to cap height. A safe default for UI and signage.
- Hind (Indian Type Foundry): slightly larger x-height, good for small sizes on screens.
- Noto Sans and Noto Sans Devanagari (Monotype for Google): the widest language coverage, with a lower Devanagari headline, so bump Hindi up about 10% when pairing with Noto Sans Latin.
- Tiro Devanagari Hindi (Tiro Typeworks): a book-style serif designed by John Hudson and Fiona Ross, for long reading and formal documents.
- Baloo 2 (Ek Type): rounded and friendly, built for display and brand use across many Indian scripts.
Display styles
For display styles the answer is more limited. A Graphic Design Stack Exchange user asked for a Devanagari match to Impact, Oswald or Anton; the answers pointed to Teko Semi Bold and Khand Bold, both free on Google Fonts.
That is typical: condensed bold Devanagari exists, but the choice is narrow, so plan brand typography with the Devanagari constraint from day one. The new commercial releases help. TypeTogether launched Bree Devanagari in March 2026, with the Devanagari led by Pooja Saxena, as an extension of its 2008 Latin Bree.

Do people read Hindi and English differently on the page?
Yes. Eye-tracking studies show readers of Hindi make longer and more frequent fixations than readers of English, and bilingual readers allocate attention between scripts according to which one they have practised most.
Layout should give the primary script the dominant position and not just alternate them mechanically.
The second wave of MECO recorded eye movements while 654 adults read comparable texts in their first language at 16 sites, including two Hindi samples: 54 readers at IIT Kanpur and 57 at IIIT Hyderabad. Across languages, first fixations lasted about 200-250 ms and reading rates ran from about 150 to 350 words per minute. Hindi readers, like Basque readers, showed longer fixations and more fixations per word, which the authors link to Hindi's rich morphology and visually complex script. The limit: the samples are university students, and per-language values sit in supplementary tables rather than the main text.
Read the paper (free)Thirty native Hindi speakers educated in English-medium schools watched short instructional videos with English or Devanagari subtitles while an EyeLink 1000 Plus tracked their gaze. With English subtitles they split attention almost evenly between faces (0.47 of dwell time) and subtitles (0.51); with Hindi subtitles, faces took 0.68-0.73 and subtitles only 0.24-0.31. Comprehension did not differ between conditions. The authors read this as attention following the more practised reading channel, while noting Devanagari's visual complexity as a rival explanation. The limit: 30 people from one UK university, one trial per condition, and English comprehension questions.
Read the paper (free)For a design system serving urban, English-medium-educated users, that suggests English may be scanned first even when Hindi is the native language, and Hindi needs size and contrast advantages to be read at all.
For a rural or Hindi-medium audience, reverse the hierarchy.
Decide the primary script per audience, not per habit.
How should a bilingual grid be structured?
Build the baseline grid on the script with the taller line, which is Devanagari, and let Latin lines sit on every grid line or every other one. If you build the grid on Latin leading, Devanagari will either overflow the module or force uneven spacing.
- Set the Devanagari body first. For example, Mukta at 17 px with a 28 px line height (1.65).
- Use the same line height for English body. Mukta's Latin at 16 px on the same 28 px grid is loose but even; or use 16 px on 24 px and accept a 4 px offset only if the two languages never sit side by side.
- Choose the layout model. Stacked (Hindi above English) suits signs and headlines. Side-by-side columns suit forms and documents, which is what the rules call "diglot" form for manuals and codes. Alternating pages suit long reports.
- Give the columns different widths if needed. The same message in Hindi often runs longer or shorter than English depending on vocabulary; measure real translated copy before fixing column widths.
- Align by headline and cap height in display type. When two headlines sit side by side, align the shirorekha with the Latin cap height, not the baselines, if the fonts differ in proportion.
The underlying maths of baseline units, modules and column widths is in the typographic grid guide. How Devanagari fonts handle conjuncts and shaping, which affects how much space a word takes, is in how Devanagari fonts work.
What do the official rules require on signage and documents?
For Central Government offices, rule 11 of the Official Languages (Use for Official Purposes of the Union) Rules, 1976 sets three requirements:
| What | Requirement |
|---|---|
| Name-plates, sign-boards, letter-heads, stationery inscriptions | Hindi and English |
| Forms and register headings | Both languages |
| Manuals, codes, procedural literature | Both, printed in diglot form |
The rules apply across India except Tamil Nadu.
Language, not typography
The rules set language, not typography: they do not specify sizes, order or fonts, and the government can exempt particular offices.
Departments add their own instructions, and state bodies often add the state language, which turns bilingual design into trilingual design. Practical consequences for designers:
- Equal treatment is implied, not measured. Matching visible size (as above) is the defensible reading; shrinking Hindi to a footnote is not.
- Order is a design decision. The rules do not fix which script comes first, so fix the order in the brand guidelines rather than per job.
- Machine translation is arriving. In October 2026, the Bhashini division and the CAG launched CAG-Bhashini Anuvaad to translate audit documents with human oversight; Bhashini says it supports 36 Indian languages for text. More bilingual documents will be produced faster, so templates need to take variable-length text.

How do you check a bilingual layout before print?
Proof with real translated copy, at real size, and with the longest words and densest conjuncts. Most bilingual failures are invisible in lorem ipsum.
- Size check. Print both scripts at final size and look at them from reading distance. If one looks heavier or larger, adjust size first, then weight.
- Matra check. Find lines with ि, ी, ु and ू near each other in consecutive lines and confirm they do not touch.
- Conjunct check. Set words with stacked conjuncts (such as द्ध or ष्ट्र) and check the font renders them correctly and in the right width.
- Weight check. Devanagari's headline adds a lot of horizontal ink, so the same nominal weight can look darker in Hindi. Drop one weight step if the Hindi block looks heavier.
- Contrast check. Thin strokes in small Devanagari lose contrast faster than Latin. Keep text colours at least at the levels in the colour contrast guide.
I switch between Hindi, Bengali, Hinglish and English myself, which is why this problem shows up in almost every brand I touch; more on how I work is on the Kala page.

What's new
- TypeTogether releases Bree Devanagari, led by Pooja Saxena, extending the 2008 Latin Bree TypeTogether, Mar 2026
- Bhashini and the CAG launch CAG-Bhashini Anuvaad for AI translation of official audit documents Indian Infrastructure, Oct 2026
- Eye-tracking study finds Hindi-English bilinguals read English subtitles far more than Devanagari ones Vision (MDPI), 2026
Questions people ask
Is there a Devanagari font that matches Impact or Oswald?
Not an exact match, but condensed bold Devanagari exists. On Graphic Design Stack Exchange the suggestions were Teko Semi Bold and Khand Bold, both free on Google Fonts. Check the headline height against your Latin cap height and adjust size so they align.
Should Hindi and English be the same font size?
Only within a harmonised family. Across families, match visible size: keep the Devanagari headline at about 1.25-1.35 times the Latin x-height, which is what most designed pairs I measured do.
What line height does Devanagari need?
Its vowel signs extend 1.06 to 1.30 em from top to bottom in the families measured here, so body text usually needs about 1.5-1.7 and headings at least 1.3 to stop matras from colliding.
Which should go first on a bilingual sign, Hindi or English?
The Official Languages Rules require both but do not fix the order or size. Follow the client's own instructions and keep the order consistent across the whole system.
Sources
- Google Design: The modern Tiro Indic collection (source for the 2012 Murty commission, the designers' research and the 2019 Google Fonts extension)
- Wikipedia: Murty Classical Library of India (dual-language facing format, January 2015 launch, Sur's Ocean, $5.2 million gift)
- Google Fonts repository: font files for Mukta, Hind, Baloo 2, Tiro Devanagari Hindi, Noto Sans Devanagari, Noto Sans, Poppins, Yatra One, Rozha One and Kalam (metrics measured Nov 2026)
- Google Fonts: Devanagari families
- Official Languages (Use for Official Purposes of the Union) Rules, 1976, as amended (CAG copy)
- Department of Official Language, Ministry of Home Affairs
- Siegelman et al.: Wave 2 of the Multilingual Eye-Movement Corpus (MECO), Scientific Data (2025)
- Romero-Ortells et al.: Subtitle Engagement Varies with Audio-Subtitle Language-Script Pairing, Vision (2026)
- TypeTogether: New release, Bree Devanagari (2026)
- Indian Infrastructure: Bhashini and CAG launch CAG-Bhashini Anuvaad (2026)
- Ek Type: multi-script type foundry
- Graphic Design Stack Exchange: Devanagari font matching Impact, Oswald and Anton
- Graphic Design Stack Exchange: Is there a non-Latin "pseudo-Latin" font?
- Graphic Design Stack Exchange: Where to find nice foreign language fonts