Supported Files & Output

What goes in, what comes out, and how long jobs are handled.

Input formats

Type Notes
.pdf Text is extracted from the PDF’s text layer. Scanned / image-only PDFs have no text layer and are skipped (see below).
.docx Text is extracted (paragraphs and headings). Original styling/layout is not preserved.
.txt Read as-is.
.md Read as-is; Markdown structure is preserved through translation.

Only files directly inside the chosen folder are processed — the app does not scan subfolders. Put everything you want translated into one folder.

Scanned / image-only PDFs

A scanned PDF is really a set of page images with no selectable text. There is nothing to extract or translate, so the app:

  1. Detects that the PDF has no usable text layer
  2. Skips it (no tokens spent)
  3. Lists it in the results as skipped, with the reason

If you need scanned documents translated, run them through OCR first to produce a text-based PDF or a .txt/.docx, then translate that.

Output layout

Each run creates a timestamped subfolder in your output folder:

Documents\BatchTranslate\
  2026-05-18_12-44-05\
    Report 2024-pdf.de-DE.md
    Report 2024-pdf.fr-FR.md
    Notes-docx.de-DE.md
    .batchtranslate-progress.json
  • One file per document per language.
  • The output name keeps the original file name, folds in the original extension (so report.pdf and report.docx never collide), and adds the language code.
  • Format is Markdown (.md) or Plain text (.txt), depending on your Step 2 choice. Original PDF/DOCX page layout and fonts are not reproduced — this is a text translator, not a layout converter.
  • .batchtranslate-progress.json is the progress checkpoint written during an active run (see below). It is safe to delete once you are happy with a finished run.

Same-language pass-through

If a target language is the same language a document is already in (as detected by the AI), the document is copied through unchanged and no tokens are spent. It appears in the results marked *“Source already in — copied unchanged (0 tokens)”*.

Languages with regional variants (English US / UK) are always translated rather than copied, so the requested variant’s localisation is actually applied.

Long documents

There is no practical page limit. Documents are split into chunks on paragraph and sentence boundaries (never mid-sentence), translated in order with terminology kept consistent across chunks, and written to disk as each chunk completes.

A .batchtranslate-progress.json file is written inside the active run folder while translation is in progress. In the current desktop UI, clicking Translate again creates a new timestamped run folder, so restarting a stopped batch starts a fresh run rather than resuming inside the previous folder.

Token size and cost

Use Estimate size in Step 2 for a rough pre-flight token range. It is deliberately approximate; once translation starts, a live projection refines it from real token usage.

The app does not show prices — providers price differently and change rates. To get a cost figure, multiply the token count by your provider’s published per-token price. See Choosing an AI Provider for rough comparisons.