In a nutshell: PDFs that contain real text are read instantly and in full. PDFs that are photos of paper ("scans") have to be read word by word, so Tresor reads the first 20 pages and tells you when it has skipped the rest.
Not all PDFs are built the same way, and the difference decides how quickly Tresor can read yours.
Kind | What it is | How Tresor reads it |
|---|---|---|
Digital | Created by software — Word, Google Docs, an exported report. The text is stored in the file. | Straight from the file. Instantly, and the whole document. |
Scanned | A photo of paper. Each page is just an image, so there is no text to read until it is worked out. | Page by page, by reading each image. Slow — roughly 1–3 seconds per page. |
Hybrid | A scan that has been made searchable. It looks like a photo of paper, but the text is in the file. | Straight from the file, exactly like a digital PDF. |
Not sure which you have? Open your PDF and try to select a line of text with your mouse.
If you can highlight and copy it, your PDF has text in it — Tresor reads it instantly.
If you can only drag a box around an area and nothing highlights, it is a scan.
[Screenshot: A PDF viewer showing a line of text being selected with the cursor, next to the same action on a scanned page where no text highlights]
When your PDF is a scan, Tresor has to read every page image individually and work out the words on it. That takes a few seconds per page, so a long scan would mean waiting minutes for an answer.
To keep things responsive, Tresor reads the first 20 scanned pages and skips the rest. If you upload a 45-page scan, your answer is based on pages 1–20. A notice in the chat tells you how many pages were skipped.
⚠️ Note: This is a temporary limit, not a rule. We're moving scanned pages onto faster, purpose-built hardware, and we expect this number to rise — and eventually disappear — as that work lands. Digital and hybrid PDFs are not affected and are always read in full.
The simplest fix. Save the pages you need as their own PDF and upload that instead. The 20-page limit then applies to those pages.
Instead of uploading the whole manual, upload just the chapter or the pages that contain what you need.
If you turn your scan into a hybrid PDF, the 20-page limit no longer applies — the text is in the file, so Tresor reads the whole document at once. This is the best option if you work with the same long document repeatedly.
You need a tool that recognises the text and saves it into the PDF. The finished file looks exactly like your scan, but you can select and search it.
🤔 Apple Preview can't do this. Preview lets you select text you can see on screen, but it never saves that text into the PDF — neither does exporting or printing to PDF. If someone tells you to "open it in Preview and save", that won't work.
Adobe Acrobat (paid) — the most direct route if you already have it:
Open your PDF and click Tools, then Scan & OCR.
Click Recognize Text, then In This File.
Save the file.
Acrobat saves the text into the PDF and leaves the pages looking the same. This is available in Acrobat Pro, not the free Acrobat Reader. Adobe's free online OCR tool does the same job if you only need it occasionally.
Free alternatives:
Google Drive — Upload the PDF, then open it with Google Docs. Drive recognises the text and gives you an editable document you can download again as a PDF. Free with a Google account; nothing to install.
NAPS2 (naps2.com) — A free app for Windows, macOS and Linux. It can import an existing PDF and add text to the pages that need it.
OCRmyPDF (ocrmypdf.com) — Free and open source, made specifically for this task. Best suited to users comfortable with a command line.
💡 Tip: Don't run these tools on a PDF that is already searchable. Doing it twice can leave two slightly different layers of text on top of each other, which makes searching and copying produce garbled results. The tools above avoid this automatically — OCRmyPDF refuses to process a file that already has text, and NAPS2 skips pages that already have it.
Digital and hybrid PDFs have no page limit. The 20-page limit applies only to pages Tresor has to read as images.
The notice tells you the exact number. When pages are skipped, the chat shows how many and links to this article.
You don't need to worry about the order. Page numbers are preserved, so answers can still point you to the right page.
Your files stay private. Making your PDF searchable happens on your own computer — Tresor never sees that step. The only exception is the Google Drive method, which uploads the file to Google first.
When you attach a PDF, Tresor opens it inside the secure enclave and checks whether each page contains text.
Pages that contain text are used directly. Pages that are images only are rendered at 144 DPI, resized to keep the file manageable, and sent one at a time to a vision model that writes out the words on the page. Those transcriptions are then stitched together with the rest of the document, keeping page numbers so answers can cite them.
Because the image pages are read one at a time, the cost grows with how many of them there are — which is exactly why the 20-page limit exists. The limit and the model used are both configuration values, so they can be changed without a code release.
Documents and images — Uploading files to conversations and projects.
— Keeping documents available across a whole project.
How Tresor protects your privacy — What happens to your files after you upload them.
