The vernacular internet is not a translation problem

Most startups in India treat vernacular content as a last step. Build the product in English, then "add Gujarati" or "add Hindi" before launch. I think that is backwards, and as someone building a product for Gujarati and Hindi speakers, I want to explain why.

The translation trap

Here is the usual sequence. A team writes the copy, the app screens and the ads in English. Then someone runs it through a translation tool, or hands it to a freelancer with a deadline. The result is grammatically correct and completely lifeless.

You have seen this. A Gujarati banner that reads like a government circular. A Hindi voice-over that uses words nobody in Ahmedabad or Indore would use at home. A sentence that is technically right and sounds like no actual person.

Readers notice instantly. They may not be able to say what is wrong, but they feel it. And the message they take away is: this was not made for me. Someone made it for somebody else and then converted it.

That is the translation trap. You have spent money to be in the customer's language, and the customer still feels like an outsider.

Language carries more than words

Translation assumes the meaning is fixed and only the words change. For most real communication, that is not true.

Take a simple shop offer. In English you might write "Festive sale, up to 30% off". A Gujarati shopkeeper speaking to a regular customer would never say it that way. They would talk about the festival by name, about what is new, about inviting the family to come and see. The persuasion is different, not just the vocabulary.

Then there is tone. Gujarati has a warmth and a kind of respectful familiarity that is very hard to carry over from English. Words like "પધારો" (padharo, please do come) do social work that "visit us" does not. Hindi has its own registers, from the formal to the street-friendly, and picking the wrong one sounds either stiff or cheap.

And there is mixing. Real people in Gujarat do not speak pure Gujarati. They mix in Hindi and English words all the time: "offer", "stock", "delivery", "free". A "correct" translation that replaces every English word with a formal Gujarati one actually sounds less natural than the way people talk.

So when I say vernacular content is not a translation problem, I mean the unit of work is not the sentence. It is the whole message: what you say, in what order, in what tone, with what mix of words.

Why voice makes it harder

Text can hide a lot. Voice hides nothing.

A written line that is slightly off might pass. The same line read aloud with the wrong stress, a mispronounced name or a flat, foreign rhythm will not. Listeners decide within a second or two whether a voice is "one of us".

This is one of the hardest parts of what Bolo Media does. Bolo makes video ads and reels with natural voices in Gujarati and Hindi, and "natural" is doing a lot of work in that sentence. Numbers, prices, place names, brand names, words borrowed from English, all of it has to sound like a local person said it. I will write separately about why Indian languages are still hard for AI voice, because it deserves its own post.

Vernacular content in India: what startups should do instead

If translation is the wrong frame, what is the right one? Here is how I think about it, especially for teams that want to reach beyond English speakers.

Start with the person, not the string. Before writing anything, ask who is hearing it, where, and what they already believe. A shop owner's customer in Surat and a college student in Delhi are not the same audience just because both understand Hindi.

Write in the language first. For Gujarati content, the first draft should be in Gujarati, by someone who thinks in Gujarati. English can be the translation, not the source.

Choose one language well over many badly. It is tempting to claim support for ten Indian languages. I would much rather do Gujarati and Hindi properly than do ten in a way that makes every one of them sound borrowed. Bolo's blog has a useful piece on how a business can decide between Gujarati or Hindi content, which is often the first real decision.

Accept code-mixing. Write the way people speak, including the English words that are now simply part of everyday Gujarati and Hindi. Purity is not the goal. Being understood and trusted is.

Test with ears, not only eyes. Read everything aloud. Better still, have someone from the target region listen. If they wince, it is wrong, whatever the dictionary says.

Why startups keep getting this wrong

I do not think it is laziness. I think it comes from how most tech teams are built.

The people writing product specs, marketing plans and pitch decks often work in English all day. English becomes the "real" version of the product, and every other language feels like a copy. Budgets follow that thinking. Localisation gets a small line at the end of the plan.

I spent many years in software teams myself, and I understand the pull. English is the language of code, of documentation, of most tools. But the customer for a lot of Indian products does not live in that world. They live on WhatsApp, on reels, in conversations at the counter, in their own language.

If your customer thinks in Gujarati, your product should too.

What I am betting on

I am building Bolo Media on the bet that the next many crores of Indian internet users will expect to be spoken to in their own language, naturally, by default. Not as a setting buried in a menu. Not as a translated afterthought.

That is a bigger idea than any one product, and I have written about how it shapes my thinking on the founder side in why I'm building Bolo Media.

I may be wrong about the details. I do not think I am wrong about the direction. The vernacular internet is not English with a different font. It is its own place, with its own tone, and the products that win there will be the ones built from inside it.