How do I OCR all files in a folder at once?

Click "Select Folder" and choose the folder with your JPG, PNG, WEBP, BMP, or PDF files — subfolders are included automatically. Pick a language, click "Start OCR All", and every file is OCR’d in your browser via Tesseract.js, several at a time, with a progress bar and per-file confidence score. When it finishes, download a ZIP of .txt files, a combined Markdown or JSONL file, a CSV, or one merged searchable PDF. No install, no upload, free.

Can I OCR a folder including its subfolders?

Yes — the folder picker reads every file inside every subfolder in one pass, and each file’s full relative path (e.g. "2024/receipts/receipt-001.jpg") is shown in the file list and preserved in the ZIP export and every combined format.

How many images or PDFs can I bulk OCR at once?

Up to 300 files and 500.0 MB total per run — a browser-memory limit, not a server one, since nothing here is uploaded. 2 parallel workers is a safe default; bump it to 4 on a fast machine for a shorter run.

Is Bulk OCR private? Does it upload my files?

Yes, 100% private. Every file is read and OCR’d in your browser tab with Tesseract.js (WebAssembly) — nothing is uploaded to Toolsda or anywhere else. You can confirm this by opening your browser’s Network tab while a run is in progress, or by disconnecting from the internet after the page (and the OCR engine) has loaded.

Can I OCR PDFs in bulk into searchable PDFs?

Yes — select a folder of scanned PDFs and each page is rendered to a canvas and OCR’d individually, then combined with "Merged searchable PDF": one PDF containing every source file’s original pixels with an invisible, word-positioned text layer over them, so Ctrl+F finds text across the whole batch.

How do I export the results of a bulk OCR run?

Six ways: a ZIP of individual .txt files (folder structure preserved), one combined Markdown file with a heading per source path, a combined JSONL file (one JSON object per file — file, path, text, tokens, confidence), a CSV, the merged searchable PDF, or copy the combined text/Markdown straight to the clipboard.

What languages does Bulk OCR support?

English, Hindi, and English + Hindi for mixed documents — the same three-option picker this site’s other OCR tools use. Tesseract.js itself supports far more languages; this list is kept deliberately small so the WASM traineddata download stays fast.

How accurate is bulk OCR, and how long does 100 files take?

Tesseract.js typically runs 90-95% confidence on a clear scan or photo and lower on a blurry or low-light one — the per-file confidence score in the table tells you which files are worth a manual check. 100 clean JPGs at 2 parallel workers usually finishes in a couple of minutes; a folder of multi-page scanned PDFs takes longer since every page is OCR’d individually.

PDF Tools

📄 Powder-Coated SteelRuns on your device

Bulk OCR — OCR All Files in a Folder at Once

Select an entire folder — JPG, PNG, WEBP, BMP, or PDF, subfolders included. Every file is OCR'd in your browser via Tesseract.js, in parallel, with a live progress bar and confidence score per file. Export one combined Markdown/JSONL/CSV file, a ZIP of individual text files, or a single merged, searchable PDF. 100% private — no upload

Input

Drop an entire folder here

JPG, PNG, WEBP, BMP, PDF · subfolders included · up to 300 files, 500 MB — nothing is uploaded

OCR Settings

Language

Parallel workers

More workers OCR more files at once — and use more memory. 2 is a safe default.

Limits

Up to 300 files, 500.0 MB total per run. JPG, PNG, WEBP, BMP, and PDF only — TIFF has no in-browser decoder, so it's skipped with a reason shown per file.

Supported

JPG · PNG · WEBP · BMP · PDF (scanned or text) · subfolders included

Combined Output

OCR results appear here as each file finishes — combined text, Markdown, JSONL, CSV, a merged searchable PDF, and a token summary.

What is bulk OCR?

Bulk OCR means running OCR (optical character recognition) on every image and PDF in a folder in one pass, instead of opening each file one at a time. Point it at a folder of 50 scanned receipts, 100 photographed invoices, or a stack of scanned contracts, and it comes back with one combined, searchable body of text — a job that takes hours by hand and a couple of minutes here.

How to OCR all images in a folder at once

Click Select Folder and choose the folder — JPG, PNG, WEBP, BMP, and PDF are all read; subfolders come along automatically.
Pick a language: English, Hindi, or English + Hindi.
Click Start OCR All. The file table updates live — status, per-file progress, and a confidence score as each one finishes.
Watch the overall progress bar (files done, elapsed time, ETA) or keep working — the combined output tabs fill in as each file finishes, not just at the end.
Download a ZIP of individual .txt files, one combined Markdown/JSONL/CSV file, or the merged searchable PDF.

No Python, no install, nothing uploaded.

Bulk OCR vs single-file OCR vs desktop tools

Method	Install?	Privacy	Searchable PDF?	Cost
Toolsda Bulk OCR	None — folder picker in browser	100% private, nothing uploaded	Yes, merged	Free
Python + pytesseract script	Python + Tesseract + pip install	Local, no UI	No — code only	Free
Desktop bulk-OCR apps	.exe/.dmg install	Local	Usually yes	Free–paid
Cloud OCR APIs (per-image billed)	Script/SDK setup	Images leave the device	Depends on service	Per image

File types and limits

JPG, JPEG, PNG, WEBP, BMP, and PDF (scanned or text) are supported — up to 300 files and 500.0 MB total in one run, which is a browser-memory ceiling rather than a server quota, since nothing is uploaded. A multi-page PDF is OCR'd page by page and combined with --- Page X/Y --- markers. TIFF is deliberately not supported: no browser can decode it through canvas without an extra library, so it's skipped with that reason shown per file rather than silently failing.

Getting the best accuracy out of a bulk run

Pick the right language — English, Hindi, or English + Hindi for mixed documents.
Higher-resolution scans OCR more accurately than a small or blurry phone photo.
Check the per-file confidence column — anything under ~80% is worth a manual look before you trust it.
2 parallel workers is the safe default; 4 finishes faster on a fast machine at the cost of more memory.
Use Retry failed rather than starting over — only the failed files re-run.

Real use cases

Receipts → expense report. OCR a folder of 100 receipt photos, download the combined text, and paste it into an LLM to categorize spend.
Scanned books → a searchable library. OCR a folder of scanned chapters into one merged searchable PDF you can Ctrl+F across.
Invoices → structured data. Export JSONL and feed it to an LLM or a script to pull out invoice numbers and totals.
Legal documents → keyword search. OCR a folder of scanned contracts into one searchable PDF and find every mention of a clause instantly.

How it works, technically

The folder picker uses the browser's webkitdirectory API (a dropped folder is read via DataTransferItem.webkitGetAsEntry(), walked recursively) — nothing is uploaded at any point. Each image is handed straight to a Tesseract.js worker; each PDF page is rendered to a canvas via pdf.js first, then OCR'd the same way. Several files run at once across a small pool of Tesseract workers rather than one at a time. The merged searchable PDF is built with pdf-lib: the original page image plus an invisible text layer positioned at each recognized word's bounding box — the same technique commercial OCR software uses, running entirely client-side.

Why a browser tool instead of a Python script?

A pytesseract loop needs Python and Tesseract installed, a script written, and produces no searchable PDF, no token count, and no UI — it's the right tool if you're already comfortable in a terminal. This tool trades that flexibility for zero setup: drop a folder, watch a progress bar, get a searchable PDF and a combined export in the format an LLM or a spreadsheet actually wants.

Related tools

Receipt OCR Parser — a folder full of receipts? Parse merchant, total, tax, and an items table into Excel per receipt instead of just raw text.
PDF to LLM Optimizer — OCR one scanned PDF and see its vision-token cost vs. its OCR'd text cost.
Image to LLM Optimizer — OCR one image and compare it against sending it as a vision tile.
File to Tokens — count tokens for any file and check whether it fits GPT-4o, Claude, or Gemini's context window.
RAG Chunker — split a large combined OCR export into overlapping chunks for retrieval.
Merge PDF and PDF to JPG — reshape files before or after a bulk run.