Pipeline

How Rebuild Page turns a PDF into Word

Generic converters write a Word file in one pass. That is fast, and it is why jackets, ID cards, and illustrated spreads come out as a flattened text dump. Rebuild Page splits the job: reconstruct the page as HTML, let you approve it, then build Word from that HTML.

The HTML is the source of truth

Layout lives in HTML you can inspect and edit. The studio shows the original page next to the reconstruction. When you change a line of text or nudge a block, the Word download is produced from the version you saved — not from a hidden second converter.

1. 01

   Upload

   Sign in and upload a PDF, JPEG, or PNG. Auto-detect chooses letter, layout, or ID unless you are calling the API with an explicit document_type.

2. 02

   Review the reconstructed HTML

   Open the job. The original is on one side and the reconstructed HTML is on the other. Edit text and layout in the studio until the page matches.

3. 03

   Download Word

   Export DOCX from that same HTML. The Word file is not a separate flattened conversion.

What auto-detect is choosing

Uploads are classified as letter, layout (illustrated spreads, jackets, artwork), or ID. Letters are mostly running text. Layout jobs keep visual structure. ID jobs keep photo, labels, and field geometry. The partner API exposes the same types as document_type: auto, letter, layout, or id.

If the page is a card-shaped scan, see ID cards to Word. If it is a jacket or illustrated spread, see illustrated PDFs. Developers can run the same pipeline from the Partner API.

Privacy and retention

Documents are not used to train AI models and are not sold. Files are stored so you can review and download, then removed after 14 days unless you keep the job. Full detail is on the Data & Privacy page.

Questions

What files can I upload?

PDF, JPEG, or PNG. One PDF at a time, or a set of page images. Rebuild Page auto-detects letters, illustrated spreads, and ID cards.

If I edit the HTML, does Word change too?

Yes. Word is generated from the HTML you approved — not from a second, independent conversion. Preview and download stay in sync.

How are credits charged?

Each job holds a per-page ceiling, then settles to measured usage. Translation is a separate hold-then-settle charge. There is no subscription.

How long do you keep the file?

Uploads and outputs are removed after 14 days unless you mark the job Keep. You can delete a job immediately from its page.