Most of what gets written about bilingual typography treats the second script as a translation problem — as if Persian were a version of English that happened to run right-to-left. The assumption, never stated, is that the Latin text is the primary one and everything else is accommodation. When we were asked to design a magazine that would publish original writing in both Persian and English, often on the same spread, we knew we could not work that way. We needed a grid that treated both scripts as native.
What we got wrong first
Our first three attempts failed in instructive ways. The first grid was a twelve-column Latin module with a Persian column count tacked on; it read beautifully in English and felt cramped and mistrusted in Persian. The second grid was the inverse, designed Persian-first; it made the English feel like it was borrowing someone else's house. The third tried to split the difference with a wider central gutter and narrower text blocks on both sides, and satisfied no one.
A grid is not a container for text. It is a set of relationships between decisions you have not made yet.
The thing we kept getting wrong was treating the grid as a finished object that text would be poured into. A grid is not a container. It is a set of relationships — between the measure, the baseline, the margin, and the hierarchy — that exist before any text arrives. Pouring two scripts into a grid designed for one will always make the second one look foreign, because the relationships were decided without it.
Starting from the measure
We threw out the column count and started from the measure — the width of a single line of body text. The comfortable measure for a humanist Latin serif at 9.5pt is around 18 to 22 words, or roughly 65 to 75 characters. For Persian, set in a nastaliq-influenced face at the same optical size, the comfortable measure is shorter — around 45 to 55 characters — because the script's horizontal connections and ascenders create more visual density per line.
So the measures could not match. But they could relate. We set the Persian measure as the base unit and made the Latin measure exactly 1.3 times that width — a ratio we arrived at empirically, by printing the same paragraph at five different ratios and asking ten people which felt balanced. One-point-three was the only ratio nobody called out as wrong.
The baseline as the unifier
If the measures had to differ, the baseline could not. We built the entire vertical rhythm on a single baseline grid — 14pt — and sized both typefaces so that their x-heights landed on the same baseline at the same leading. This is the part most designers skip, because it is fiddly and unglamorous. It is also the part that makes a bilingual spread read as one design instead of two pasted together.
The Persian body is set 8% larger in point size than the Latin, to compensate for the optical weight of the script. At a glance, both blocks of text appear to be the same size. This is the goal. Optical equivalence, not numerical equivalence, is what lets a reader's eye move between the two without recalibrating.
Optical equivalence, not numerical equivalence. The reader's eye does not know what a point is.
What it cost
Building a grid this way takes about three times as long as building one for a single script. The trade-off is that once it exists, every subsequent decision is faster, because the relationships are already decided. We have used variants of this grid on four projects since, and each time the setup has taken a week instead of a month.
The magazine, when it shipped, read as one thing. Reviewers described it as bilingual without describing it as a translation, which is the highest compliment we could have asked for. The grid was invisible, which means it was doing its job.