Writing about sex
in Hindi.

The obvious way to make an app bilingual is to write it in English and run it through a translator. For most products that is fine. For this one it fails at the first sentence, and it fails in a way that is invisible to whoever commissioned the translation.

What follows is what we found writing 1,073 pieces in Hindi instead of translating them.

Three registers, and none of them is the one you want

Hindi gives you at least three vocabularies for this subject, and each one carries a room with it.

There is the Sanskritic register — the words a textbook or a scripture would use. It is precise and it is completely unembarrassed, which sounds ideal until you use it between two people in a bedroom, where it lands like a doctor has walked in. It is the register of instruction, not of wanting.

There is the Perso-Arabic register that Hindi shares with Urdu, and it is the most beautiful of the three. It is also the register of poetry and film songs, which means it arrives pre-loaded with longing and distance. It is very good at wanting somebody who is not there. It is oddly bad at somebody who is in the room.

And there is the colloquial register, which is where the actual words live, and where a great deal of it is abusive by construction. Much of the everyday sexual vocabulary in Hindi exists primarily as insult. You cannot simply lift it into a sentence meant to be tender.

Every available word arrives
carrying a room with it.

What people actually do

They code-switch, and they do it precisely.

Listen to how a couple in Delhi or Mumbai actually talks about this and you will hear Hindi carrying the feeling and English carrying the anatomy — often in the same sentence. That is not laziness and it is not aspiration. The English word is doing a specific job: it is lighter. It has no childhood attached to it, no relative who used it, no scripture and no insult. It is a word you can say without the word saying something about you.

A translator will not reproduce that, because a translator is trying to render meaning and this is not a meaning effect. It is a weight effect. Ask for a Hindi version of an English sentence and you will get every word replaced, including the four that were in English on purpose.

So the Hindi in OOHLA is not the English translated. It was written by deciding, phrase by phrase, which register the sentence needed and where a borrowed word was doing something a native one could not.

Then the typography goes wrong

Suppose you get the words right. Now put them into an interface that was designed in Latin script, and watch four things break at once.

There is no uppercase. Devanagari has no case distinction at all. Every small label in a modern interface — the eyebrow above a heading, the tiny category tag — is uppercase Latin by convention. In Devanagari that convention has nothing to apply to, so a naive port simply loses a whole level of visual hierarchy and nobody notices, because the text is still legible.

Letter-spacing is destructive. That same eyebrow usually carries wide tracking. In Devanagari, letters join into conjuncts and vowel marks sit above and below the line, attached to the letter they belong to. Adding space between characters does not look airy; it detaches marks from their letters. It is closer to breaking the word than to styling it.

Italic does not exist. There is no true italic Devanagari in the fonts anyone actually has. Ask a browser for one and it will synthesise it by slanting the upright — which distorts the horizontal headline stroke that the whole script hangs from, and produces something that reads as a rendering fault rather than as emphasis.

Small sizes cost more. A 9px Latin label is hard to read. A 9px Devanagari label loses the matras — the marks that decide which vowel you are looking at — so it is not small, it is ambiguous.

What we did about it

Every one of those is a rule in the stylesheet rather than a thing a writer has to remember. When the interface is in Hindi: tracking goes to normal, uppercase is switched off, synthesised italics are refused, the monospace stack is dropped entirely rather than extended, there is a floor under the small print, and line height opens up because the script needs vertical room for its marks.

None of that is visible when it is working. It is only visible when it is missing, and then it reads to a Hindi speaker as an app that was clearly not built for them — which is a judgement made in about four seconds and very rarely revised.

Vieena answers, rather than answers-then-translates

The same argument applies to the part of the product that writes new sentences. Vieena is told the language the reader is reading in and answers in it. The alternative — answer in English, translate afterwards — is the failure above, at runtime, forever.

A sex educator’s register is the whole job. How direct to be, how clinical, which word for which part of the body, when to be plain and when to be gentle. That is exactly what does not survive a round trip through a general-purpose translator, and this is the one room where getting the register wrong is not a rough edge. It is the failure.

Translated intimacy reads as
intimacy about somebody else.

The part that is not solved

Hindi is not one thing. What is comfortable in one city is coarse in another, and a good deal of what we wrote will land differently across a very large country. We have written one register and it is a choice, not a neutral default.

And two languages is two, out of a great many that people in India actually make love in. Every other one is unaddressed, and calling that finished would be its own kind of dishonesty.