Convert an Excel Key-Value Sheet to i18n JSON Translation Files

Flat JSON from a two-language sheet

The default case: a key column and one value column. You name the value column you want and get a flat object. To also produce the English file, run again with valueColumn: English.

Sheet (first tab):
key          | English   | French
-------------|-----------|----------
greeting     | Hello     | Bonjour
farewell     | Goodbye   | Au revoir

Options: keyColumn=key, valueColumn=French, nested=false

Output → french.json:
{
  "greeting": "Bonjour",
  "farewell": "Au revoir"
}

Nested object from dotted keys

Turn nested on and dotted keys expand into a tree. Keys that share a prefix merge under the same parent object, which is what most modern i18n libraries expect.

Sheet:
key            | French
---------------|------------
nav.home       | Accueil
nav.settings   | Paramètres
footer.legal   | Mentions légales

Options: keyColumn=key, valueColumn=French, nested=true

Output → french.json:
{
  "nav": {
    "home": "Accueil",
    "settings": "Paramètres"
  },
  "footer": {
    "legal": "Mentions légales"
  }
}

Blank separator rows are skipped

Translators often leave an empty row between sections for readability. Because empty keys are skipped, those rows simply vanish from the output — no "": "" entry, no error.

Sheet:
key         | French
------------|----------
login       | Connexion
            |            ← blank separator row
signup      | Inscription

Options: keyColumn=key, valueColumn=French

Output → french.json:
{
  "login": "Connexion",
  "signup": "Inscription"
}
(input rows: 3, output keys: 2)

Building three locale files from one sheet

There is no single-click multi-language export. You run the tool once per language column, changing only valueColumn. The key structure is identical across all three, which is exactly what you want for parallel locale files.

One sheet, three runs:
key        | en       | fr        | de
-----------|----------|-----------|----------
save       | Save     | Enregistrer | Speichern
cancel     | Cancel   | Annuler   | Abbrechen

Run 1: valueColumn=en  → en.json
Run 2: valueColumn=fr  → fr.json
Run 3: valueColumn=de  → de.json

Drop each into your /locales folder.

Values keep their exact whitespace and characters

Only the KEY is trimmed. The value is written verbatim — leading spaces, interpolation placeholders, and ICU braces all survive untouched, so your {count} and {{name}} placeholders are preserved.

Sheet:
key            | French
---------------|----------------------------
welcome.user   | Bonjour, {name} !
items.count    | Vous avez {count} articles
indented.note  |   Texte avec espaces

Output → french.json:
{
  "welcome.user": "Bonjour, {name} !",
  "items.count": "Vous avez {count} articles",
  "indented.note": "   Texte avec espaces"
}
(value whitespace preserved exactly)

Keys are written into a single object, so if greeting appears on two rows, the second row's value overwrites the first — silently, with no warning. The output key count will be lower than the input row count. If you expected both, the spreadsheet has an accidental duplicate; dedupe it first with the duplicate-purge tool before generating.

Matching is by exact header text. If you set valueColumn: french but the header is French, every cell reads as empty string, and you get keys mapped to "". Check the header capitalisation and spelling against the first row of the sheet.

A row whose key cell is empty (or whitespace-only, which trims to empty) is skipped entirely — neither key nor value reaches the output. This is intentional so separator rows don't pollute the JSON, but it also means a row where the translator forgot the key is dropped without notice.

With nested: false, a key like home.title stays a literal flat string key "home.title". This is valid and fine for libraries that accept flat dotted keys (some i18next setups). It only becomes a problem if your library expects nesting — then turn nested on.

With nested: true, if menu has a value AND menu.home also exists, they collide on the menu path. The nesting logic overwrites a string with an object (or vice versa) depending on row order, so one of them is lost. Rename the conflicting key — menu and menu.home can't coexist in a nested tree.

Only the first sheet is parsed. If your workbook keeps notes on sheet 1 and translations on sheet 2, the tool reads the notes. Move the translation table to the first sheet, or save that sheet as a standalone CSV before uploading.

Cell values are converted with String(...). A numeric 0.5 becomes the string "0.5"; a date cell becomes its displayed string form. JSON values are always strings here — if you need typed values you'll cast them in your app, not in this tool.

A Free-tier file over 5 MB or 10,000 rows is rejected before processing. Split the sheet, trim to one language column, or upgrade. Locale files rarely approach these limits, but a master sheet with thousands of keys across dozens of languages can.

Output is built with JSON.stringify, so a value containing a double quote or backslash is escaped correctly (\", \\). You never need to pre-escape the spreadsheet — that is the whole point of generating rather than hand-editing JSON.

An untranslated row (key present, value blank) writes "key": "". The key is kept because it has a key; only the value is empty. This is useful as a checklist of untranslated strings — search the output for : "" to find gaps.