How to Translate a Birth Certificate to English

The page below is a Mexican birth certificate written in Spanish. We uploaded the one-page scan to LangTo and chose English.

The translation took five and a half minutes.

The Spanish source is on the left. It is a sample built to the official unified format used in Mexico. The English translation is on the right.

The translated page reads in English without making the reader learn a new layout. “Acta de Nacimiento” becomes “Birth Certificate.” The section headings, field labels, certification paragraph and verification footer are translated in place. The registered person and both parents remain under the same headings, and the date of birth reads 17 MARCH 1994 instead of 17 DE MARZO DE 1994.

The names, registry identifiers, barcodes, QR codes and signature still come directly from the source scan.

Translate only what needs translation

A birth certificate leaves little room for a plausible-looking mistake. One missing digit in an identifier or one altered letter in a surname can make an otherwise fluent translation a long checking job.

LangTo handles the scanned page by replacing only the areas that contain language to translate. The rest of the source image remains in the finished PDF. On this page, the English field labels and legal text were newly set as selectable text. The names of the registered person and both parents were not retyped. Neither were the three 18-character CURP identity codes, the 20-digit electronic identifier, the folio and validation numbers, the barcodes, the QR codes or the registrar's signature.

That means zero retyping for those regions. The translation cannot drop a digit from an identifier it never rewrites, and it has no occasion to turn MONTAÑO into MONTANO. The finished PDF continues to carry the same source content.

We rendered both files at 300 DPI and compared those regions, the sample stamp included. They match the source scan pixel for pixel, apart from one row of pixels along the top edge of the first CURP box, next to the label that was replaced above it.

1, a CURP as the scan has it. 2, each parent's values under their own headings. 3, the date of birth in English. 4, the certification paragraph, translated in its box. 5 and 6, the QR code, validation number, barcode and signature, all from the scan.

A date may need a different written form even though its value stays unchanged. 17 DE MARZO DE 1994 therefore becomes 17 MARCH 1994; the day itself stays the same. The same distinction applies throughout document translation: an identifier preserves its characters, while a date may change form to remain unambiguous in the translation language. We explain that distinction in more detail in our article about numbers in PDF translation.

Fit translated text to the original page

A translation rarely takes exactly the same space as its source. A PDF translator therefore has to decide what happens when a field label grows, a sentence becomes shorter or a legal paragraph needs another line.

LangTo starts from the source text's position, alignment, type size, line spacing and surrounding boundaries. It keeps those measurements where the translated text fits. When the translation is longer, it uses the space around the text first, then the spacing, and reduces the type size last. Headings remain centered, labels remain with their values, and paragraph text stays inside its original box. Text that still does not fit is never left to spill over a neighboring field: the translator words it more tightly or accepts a smaller size, without dropping anything the source says.

The small certification paragraph on this certificate states the legal basis for the certified copy, describes the electronic and handwritten signatures, and gives the issue date.

On the delivered certificate it is selectable English text, set inside its box beside the marginal notes. A los 12 días del mes de mayo de 2026 reads On 12 May 2026, and Doy fe reads I hereby certify.

The paragraph on the scan above the English text in the delivered PDF.

The rest of the page follows the same rule. English replaces the Spanish title, table headings and footer. The green security frame, data bands, seal, signature block and machine-readable areas keep their positions. An English reader can follow the record, while someone comparing it with the source still finds each field in the same place. The mechanics of translating a scanned PDF while keeping its layout are covered separately.

Use first-tier models, then keep upgrading

Translation quality still depends on understanding what every part of the page means. The model has to recognize that Acta de Nacimiento names the document, that Datos de Filiación introduces the parents' details, and that the certification paragraph is legal prose rather than a collection of labels.

LangTo uses current first-tier multimodal reasoning models for production translation. The model sees the page and works with PDF inspection and layout tools, so it can handle the language and the page as one task instead of translating extracted sentences without their visual context.

The product is not permanently tied to a model name. A new model is tested on a fixed set of real documents, and its output is compared with the current one by reviewers who are not told which is which. A stronger model enters production only after it proves that it can deliver those files reliably. That policy lets the translation ceiling rise as the field improves without asking users to gamble on an untested release.

Review the places most likely to need attention

Before the page is delivered, a second AI model from a different provider reviews it against the scan. It checks whether the meaning changed, whether numbers, dates, names and identifiers match the original, and whether anything was left untranslated. A place that only the original can settle, such as a character the scan does not make clear, is marked on the translated page itself, and the top of the result page shows how many such places there are.

The delivered certificate marked no place for checking. It carries two translator's notes, which name the CIVIL REGISTRY watermark and the SAMPLE stamp.

On a certificate, the three CURP codes, the validation number and the year, book and entry numbers deserve a direct comparison with the scan whatever the marks say. A recipient who looks up the record uses those characters, so one wrong character is enough to make the record unfindable.

Every piece of translated text is editable online. Click a field label, a heading or a sentence of the certification paragraph, type the change, and the page is laid out again around it.

Try it on your own certificate

Every PDF brings its own scan quality, stamps, shadows, handwriting and subject matter. The most useful evidence is therefore the result from the actual page in hand.

Every LangTo account starts with 5 free pages, with no watermark and no card needed, so a one-page birth certificate can be translated, edited and downloaded without paying anything. After that you buy pages, at $0.90 a page with a minimum of 10, or subscribe. If a translation can't be delivered, its pages come back to your balance.

Your documents are never used to train AI models. You can delete one from your account at any time, and the original and translated files are deleted 30 days after upload.

Translate your birth certificate