How Long Would It Take to Translate the Renaissance?

This is the third time we have tried to put a number on the same problem. First we counted how much of the Renaissance has been translated into English (almost none of it). Then we measured how many new translations appear each year (fewer than anyone guesses). That second post ended on a question we did not answer: at the current rate, how long would it take to finish? This is the division.

The short answer, for Latin alone, is about ten thousand years.

The rate

To know how long the work would take, you first need to know how fast it goes. We have a unified catalogue of 13,862 English translations of pre-modern works, drawn from the UNESCO Index Translationum, the Library of Congress, the Loeb and I Tatti series, and about forty other sources. At its best-documented stretch, 2000 to 2008, roughly 145 unique translations appeared each year across every source we track. Of those, about 72 were first translations of a work that had never been in English before; the rest were new versions of things already available, the forty-first Aeneid.

But those 72 are first translations of all pre-modern Latin and Greek, most of it classical and medieval. The number that appear from the Renaissance window specifically, 1450 to 1700, is smaller: our best estimate is 20 to 40 a year at peak, and fewer now that UNESCO stopped collecting in 2009. For a hard external anchor, Harvard’s I Tatti Renaissance Library, the single most ambitious effort ever mounted to translate Renaissance Latin, has produced about 100 volumes since 2001, or roughly four a year.

So the rate is somewhere between four a year, for the organised, funded, flagship effort, and forty a year, for everyone on earth combined at the most productive moment we can document. Call it a few dozen. It is not rising.

What is left

The Universal Short Title Catalogue records 499,607 Latin editions printed in Europe between 1450 and 1700, which cluster into roughly 362,000 distinct works. How many have an English translation?

This is where you have to be careful, and we have learned to be. If you flag every Latin edition written by an author who has any translation to his name, about 13 per cent of editions light up. But a hand-audit shows that flag is only about 70 per cent precise at the level of the individual work, and badly unit-sensitive: a translated author drags along his untranslated minor works (Grotius’s wedding poems, Erasmus’s school grammar), and prolific famous authors have far more editions than obscure ones, so counting by edition over-represents the translated. Count by work instead, and the figure collapses to one to two per cent. About 98 per cent of the Latin works of the early modern period — on the order of 355,000 of them — have never been translated into English at all.

Much of that is exactly what you would expect to find at the bottom of a print archive: almanacs, funeral orations, dissertations, papal bulls, occasional poems for weddings and coronations. But the same audit keeps turning up real books too, by real authors, that simply no one has ever carried into English. The bulk is genuinely dark.

The division

355,000 untranslated works, divided by a few dozen new translations a year. The arithmetic is brutal and not very sensitive to the assumptions you choose.

The central estimate is about ten thousand years. Push every assumption in the optimistic direction at once, and assume our catalogue undercounts the real translation rate by a factor of three, and that half of the untranslated works are trivial ephemera you would never bother with, and the answer is still over two thousand years. There is no honest set of numbers that brings it under a millennium.

Ten thousand years

It helps to hold that figure against some others. Writing itself is about five thousand years old. The printing press is not yet six hundred. The Loeb Classical Library, the longest continuously running translation series in existence, has produced about 540 volumes in its 115 years. At the Loeb’s lifetime pace, the Latin Renaissance would outlast the species that started it. We are not behind on this work. We were never going to finish it. The gap was not a backlog; it was a permanent condition.

And this is Latin only. It leaves out Greek, and the larger manuscript tradition before print, and the eighteenth century, which probably produced more books than the previous 250 years combined. It leaves out the entire written inheritance of China, India, the Islamic world, and everywhere else, all of it translated at similar rates or worse. The ten thousand years is a floor on a corner of the problem.

What changed

We ran these numbers because we are watching the constant in them break. The reason the gap never closed is that the cost of a translation was fixed at a human scale: months of a skilled person’s life per book, and only so many such people. That is the budget that produced ten thousand years. Machine translation has not improved that budget. It has removed it. A current model reads a page of seventeenth-century Latin and returns a serviceable English draft in seconds, for a fraction of a cent.

We have spent the past few years building a library on that fact. It now holds more than 15,000 historical books translated into English, thousands of them Latin, each shown beside the original page. What the table above frames as a hundred centuries of work is, at machine speed, a matter of years. That is the whole of the claim, and it is enough.

It is not a claim that the machine replaces the scholar. A fast translation is not a true one, and a model that has learned the music of a language can perform it over an emptiness; the original has to stay beside the translation, and the reading and the checking and the judgment are still entirely human, and there is now an almost unimaginable amount of them to be done. But the bottleneck that fixed the budget for five hundred years is gone. The question was never whether the Renaissance was worth translating. It was whether anyone could afford the time. For five centuries the answer was no. It has just become yes, and a harder question takes its place: not whether we can translate it, but whether we can read it.

Sources: Universal Short Title Catalogue (Latin editions 1450–1700, queried 31 May 2026); Source Library translation catalogue and rate analysis; manual QC audit of the Latin author-match flag, 31 May 2026. The work-level translation rate is an estimate with real error bars; the order of magnitude is robust, the precise figure is not.