Pull tables out of a PDF with position-based table detection, review and edit them, then export to CSV, Excel, JSON, XML, HTML or Markdown. Detects multiple tables per page, supports page ranges and optional OCR (English & Hindi) for scanned PDFs. Everything runs in your browser; your file is never uploaded.
📤 Drag a PDF file here or click to choose
Rows are grouped by vertical position and columns are found from the whitespace between them, so even right-aligned numbers line up.
A page is split into regions at blank bands, so several separate tables on one page are detected as separate tables.
A table that continues over many pages — like a bank statement — is stitched into one continuous sheet, with the repeated headers removed automatically.
Amounts, including values with commas and currency symbols, can be exported as real numbers so you can total and sort them straight away.
Review the detected table and fix anything: edit any cell, add or delete rows and columns, and remove empty ones before exporting.
Download as CSV, Excel (XLSX/XLS/ODS), JSON, XML, HTML or Markdown — whichever suits your next step.
Extract from one page, a custom range like 1-3,5, or every page in the document at once.
Optionally treat the first row as a header — used for JSON keys, the Markdown header row and Excel styling.
For image-only PDFs, built-in optical character recognition reads the table text in English, Hindi or both — no AI service.
Search within an extracted table to quickly find and highlight the rows you need before exporting.
See the PDF page next to the detected tables so you can compare and confirm before you export.
Detection, editing and export run locally with pdf.js, SheetJS and Tesseract. Your PDF is never uploaded — no server, no account.
The ApneSoftware Extract Table from PDF tool detects the tables inside a PDF, lets you review and edit them, and exports them to whatever format you need — CSV, Excel, JSON, XML, HTML or Markdown — entirely inside your browser and without uploading the file anywhere. PDF is a fixed, final-form format designed for faithful display, not for data. The moment you need the data that a table holds — to total a column, sort a list, import it into a database, feed it to a script, or paste it into a spreadsheet — you hit a wall, because copy-pasting a table out of a PDF usually collapses a neat grid into a jumble of runs. This tool bridges that gap: it reads the text out of your PDF together with the exact position of every fragment, reconstructs the rows and columns from that geometry, and hands you a real, editable table you can correct and then export cleanly.
At its core the tool uses pdf.js, the same engine that renders PDFs inside Firefox, to read each selected page and pull out every piece of text along with its coordinates, width and font. From that raw data it rebuilds the tables in a few stages. First it groups fragments into rows by their vertical position, using a tolerance derived from the text size so a single visual line is not split in two. Then it works out the columns: rather than guessing column breaks line by line — which makes columns drift — it looks at the horizontal whitespace between the text of the data rows and places a column boundary at every consistent gap, so columns line up straight down the page and right-aligned numbers group correctly. Crucially, it also segments each page into regions: where a large blank band separates two blocks of content, they are treated as separate tables, each with its own columns. That means a page holding, say, a transaction table and a summary box below it yields two correctly-structured tables instead of one mangled grid.
Real documents rarely give up their structure easily, so the tool is built to be honest and controllable rather than a black box. Once a page is processed you see each detected table rendered as an editable grid right next to a picture of the PDF page, so you can compare them. Detection is very good for clearly-spaced tables — the columnar reports, bank and card statements, price lists, invoices and schedules that make up the vast majority of real needs — but no purely automatic method is perfect on every layout, which is why editing is a first-class feature here. You can click into any cell and correct it, add or delete rows and columns, and strip out empty rows and columns with a single option. This review-and-fix step, inspired by desktop tools like Tabula and the table tools in Adobe Acrobat, Foxit and Nitro, is what turns a good-enough automatic guess into exactly the table you want — and because it all happens locally and instantly, it is fast.
Export is deliberately broad because different jobs need different shapes. CSV is the universal choice that imports into virtually any application. Excel output comes in the modern .xlsx workbook (with a bold, shaded header row and auto-fitted columns), the older .xls binary for legacy software, and the .ods OpenDocument format for LibreOffice and OpenOffice. For developers, JSON turns the table into an array of records (using the first row as keys when you mark it as a header) and XML produces a clean, structured document. For the web and documentation, HTML gives you a ready-to-embed table and Markdown produces a pipe table that drops straight into a README, a wiki or a GitHub comment. You can also simply copy the table as tab-separated text and paste it directly into Excel or Google Sheets. When you extract from several pages, the export bundles every detected table together — separate sheets in an Excel workbook, or one file per table where the format calls for it.
Not every PDF contains real, selectable text. Many documents — especially scanned statements, receipts, forms and older records — are actually images of pages with no text layer underneath, and a plain extractor produces nothing from them. To handle these, the tool includes traditional optical character recognition (OCR) powered by the open-source Tesseract engine. When you enable OCR, each scanned page is rendered to an image and read, and the recognised words are grouped back into rows and columns by their positions, so even a scanned table becomes a usable, editable grid. You can choose the recognition language — English, Hindi, or both together — which is essential for Indian documents, and the tool detects on upload whether a PDF looks scanned and suggests turning OCR on. This is classic pattern-matching OCR that runs on your own device, not a generative AI service.
The whole experience is designed to feel like the professional tools people already know while remaining completely free and private. When you add a file the tool reports its page count and whether it contains selectable text; a progress bar shows extraction advancing page by page, which matters especially for OCR; a cancel button lets you stop a long job; and a search box inside each table lets you find and highlight rows before exporting. You can extract from a single page, a custom range like 1-3,5, or the whole document. Throughout, the interface adapts to phones and tablets, the tables scroll on small screens, and the status line explains clearly what happened — including a helpful message if a page turns out to be a scan and OCR was not enabled.
It is worth being realistic about what any PDF table extractor can and cannot do, because it saves a lot of frustration. A PDF stores positioned glyphs, not a logical table: it knows a character sits at a point on the page, but it does not necessarily record that a group of numbers forms a table, which cells are merged, or where one column ends and the next begins. Rebuilding the table means inferring that lost structure from the geometry, and that inference is excellent for clearly-spaced tables but imperfect for heavily designed pages with merged cells, nested tables, or borderless multi-column layouts, which may need a quick tidy-up using the editing tools. This tool is deliberately tuned for the common, high-value case: getting clean, aligned data out of ordinary tables so you can use it — and where automation falls short, the editing step lets you finish the job by hand.
Above all, this tool is private by design. Every step — reading the PDF, detecting the tables, recognising text with OCR, and building the export file — happens entirely on your own device using pdf.js, SheetJS and Tesseract. Nothing is uploaded to a server, nothing is stored, and no account or email is required; after the libraries have loaded you can even work offline. This matters enormously for a data tool, because the PDFs people extract tables from are so often sensitive — bank and card statements, invoices, payroll, tax and medical documents, official records — and uploading them to an online extractor, as many free services quietly do, would expose exactly the data that ought to stay private. Here, that never happens. Whether you are an accountant pulling figures from statements, an analyst gathering data from reports, a developer wiring PDF tables into a pipeline, a researcher compiling numbers, or anyone who needs the data out of a PDF, this tool gives you fast, structure-aware, private table extraction with nothing to install.
ApneSoftware Extract Table from PDF टूल किसी PDF के अंदर मौजूद टेबल को पहचानता है, आपको उन्हें देखने और संपादित करने देता है, और जिस भी प्रारूप की ज़रूरत हो उसमें निर्यात करता है — CSV, Excel, JSON, XML, HTML या Markdown — पूरी तरह आपके ब्राउज़र के अंदर और फ़ाइल को कहीं अपलोड किए बिना। PDF एक स्थिर, अंतिम‑रूप प्रारूप है जो दिखाने के लिए बना है, डेटा के लिए नहीं। जैसे ही आपको टेबल का डेटा चाहिए — किसी कॉलम का योग, सूची क्रमबद्ध करना, डेटाबेस में इम्पोर्ट करना, किसी स्क्रिप्ट को देना, या स्प्रेडशीट में पेस्ट करना — आप एक दीवार से टकराते हैं, क्योंकि PDF से टेबल कॉपी‑पेस्ट करने पर अक्सर सुंदर ग्रिड बिखर जाता है। यह टूल उस खाई को पाटता है: यह टेक्स्ट को हर टुकड़े की सटीक स्थिति के साथ पढ़ता है, उसी ज्यामिति से रो और कॉलम पुनर्निर्मित करता है, और आपको एक असली, संपादन‑योग्य टेबल देता है।
मूल रूप से यह टूल pdf.js का उपयोग करता है — वही इंजन जो Firefox में PDF रेंडर करता है — हर चुने पेज को पढ़कर हर टेक्स्ट टुकड़े को उसके निर्देशांक, चौड़ाई और फ़ॉन्ट के साथ निकालने के लिए। उस डेटा से यह टेबल कई चरणों में बनाता है। पहले यह टुकड़ों को उनकी ऊर्ध्वाधर स्थिति के आधार पर रो में समूहित करता है। फिर यह कॉलम तय करता है: हर पंक्ति पर अलग अनुमान लगाने के बजाय, यह डेटा rows के टेक्स्ट के बीच की खाली जगह देखता है और हर लगातार अंतराल पर कॉलम‑सीमा रखता है, ताकि कॉलम सीधे रहें और right‑aligned संख्याएँ सही समूहित हों। सबसे ज़रूरी — यह हर पेज को regions में बाँटता है: जहाँ बड़ी खाली पट्टी दो हिस्सों को अलग करती है, उन्हें अलग टेबल माना जाता है।
असली दस्तावेज़ आसानी से अपनी संरचना नहीं देते, इसलिए यह टूल black box नहीं, बल्कि ईमानदार और नियंत्रण‑योग्य बनाया गया है। पेज process होने पर हर पहचानी टेबल एक editable ग्रिड के रूप में PDF पेज की तस्वीर के बगल में दिखती है। पहचान स्पष्ट‑स्थान वाली टेबलों के लिए बहुत अच्छी है — रिपोर्ट, बैंक व कार्ड स्टेटमेंट, price lists, इनवॉइस — पर कोई भी स्वचालित तरीका हर layout पर परफ़ेक्ट नहीं होता, इसलिए यहाँ संपादन एक प्रमुख सुविधा है। आप किसी भी cell में क्लिक करके सुधार सकते हैं, rows और columns जोड़/हटा सकते हैं, और खाली rows/columns एक विकल्प से हटा सकते हैं। यह review‑and‑fix चरण — Tabula, Adobe Acrobat, Foxit और Nitro जैसे टूल से प्रेरित — अच्छे‑खासे अनुमान को बिल्कुल वही टेबल बना देता है जो आप चाहते हैं।
निर्यात जानबूझकर व्यापक रखा गया है क्योंकि अलग कामों को अलग आकार चाहिए। CSV सार्वभौमिक विकल्प है। Excel आउटपुट आधुनिक .xlsx (bold, shaded header और auto‑fit columns के साथ), पुराना .xls, और LibreOffice के लिए .ods में मिलता है। डेवलपर्स के लिए JSON टेबल को records की array में बदलता है (header चिह्नित करने पर पहली row keys बनती है) और XML एक साफ़ structured दस्तावेज़ देता है। वेब व documentation के लिए HTML तैयार table देता है और Markdown एक pipe table बनाता है जो README या GitHub comment में सीधे चलता है। आप टेबल को tab‑separated text के रूप में कॉपी करके Excel या Google Sheets में पेस्ट भी कर सकते हैं।
हर PDF में असली, चयन‑योग्य टेक्स्ट नहीं होता। बहुत से दस्तावेज़ — खासकर स्कैन किए स्टेटमेंट, रसीदें, फ़ॉर्म और पुराने रिकॉर्ड — असल में पेजों की तस्वीरें होती हैं। इन्हें संभालने के लिए, टूल में ओपन‑सोर्स Tesseract द्वारा संचालित पारंपरिक OCR शामिल है। OCR चालू करने पर हर स्कैन पेज एक छवि में रेंडर होकर पढ़ा जाता है, और पहचाने शब्द उनकी स्थिति से rows और columns में वापस समूहित होते हैं। आप भाषा — अंग्रेज़ी, हिंदी, या दोनों — चुन सकते हैं। यह पारंपरिक OCR है जो आपके डिवाइस पर चलता है, कोई AI सेवा नहीं।
पूरा अनुभव पेशेवर टूल जैसा बनाया गया है, फिर भी पूरी तरह मुफ़्त और निजी। फ़ाइल जोड़ने पर टूल पेज संख्या और यह बताता है कि टेक्स्ट चयन‑योग्य है या नहीं; एक प्रगति पट्टी extraction को पेज‑दर‑पेज दिखाती है; एक कैंसल बटन लंबे काम को रोकता है; और हर टेबल में एक search box rows खोजने व highlight करने देता है। आप एक पेज, 1-3,5 जैसी custom range, या पूरा दस्तावेज़ extract कर सकते हैं। पूरे समय interface फ़ोन और टैबलेट के अनुसार ढलता है।
यह वास्तविक होना उपयोगी है कि कोई भी PDF table extractor क्या कर सकता है और क्या नहीं। PDF स्थित ग्लिफ़ संग्रहीत करता है, तार्किक टेबल नहीं। टेबल फिर से बनाने का अर्थ है उस खोई संरचना का ज्यामिति से अनुमान लगाना, और यह अनुमान स्पष्ट‑स्थान वाली टेबलों के लिए उत्कृष्ट है पर merged cells, nested tables या borderless बहु‑कॉलम layout वाले पेजों के लिए अपूर्ण — जिन्हें editing tools से जल्दी ठीक किया जा सकता है।
सबसे बढ़कर, यह टूल डिज़ाइन से ही निजी है। हर चरण — PDF पढ़ना, टेबल पहचानना, OCR, और export फ़ाइल बनाना — पूरी तरह आपके डिवाइस पर pdf.js, SheetJS और Tesseract से होता है। कुछ भी सर्वर पर अपलोड नहीं होता, कुछ भी सेव नहीं होता, और किसी अकाउंट की ज़रूरत नहीं। यह किसी डेटा टूल के लिए बहुत मायने रखता है, क्योंकि जिन PDF से लोग टेबल निकालते हैं वे अक्सर संवेदनशील होती हैं — बैंक व कार्ड स्टेटमेंट, इनवॉइस, वेतन, कर व मेडिकल दस्तावेज़। यहाँ वह कभी नहीं होता। चाहे आप अकाउंटेंट हों, विश्लेषक, डेवलपर, शोधकर्ता, या कोई भी जिसे PDF से डेटा चाहिए — यह टूल आपको तेज़, संरचना‑जागरूक, निजी table extraction देता है, बिना कुछ इंस्टॉल किए।
This tool extracts tables entirely inside your browser using pdf.js, SheetJS and (for scanned files) the Tesseract OCR engine. Your file is never uploaded to any server, never stored, and never transmitted anywhere. No account, email or signup is required, and once the libraries have loaded you can work offline. Because the PDFs people extract tables from — bank and card statements, invoices, payroll, tax and medical documents — frequently contain sensitive financial and personal data, this local-only design keeps them under your sole control from upload to export.
Works in all modern browsers — Google Chrome, Mozilla Firefox, Microsoft Edge, Safari, Brave and Opera — on Windows, macOS, Linux, Android and iOS. It relies only on JavaScript and the standard File, Blob and Canvas APIs. On phones and tablets the layout stacks into a single column and the tables scroll horizontally, with touch-friendly controls. OCR downloads a language model once (a few megabytes) and then works offline for that language.
Text PDFs (with selectable, machine-readable text) extract directly and most accurately — including Hindi and other Unicode text. Scanned / image-only PDFs have no text layer; enable OCR to read the table text from the page images (English, Hindi or both). Mixed PDFs that contain both are handled page by page. Password-protected PDFs must be unlocked first with our Unlock PDF tool.
CSV — a universal plain-text table that imports into almost anything. Excel (.xlsx) — a styled, editable workbook; tables that repeat across pages are stitched into one sheet by default (or one sheet per table). Excel 97–2003 (.xls) and ODS — for legacy software and LibreOffice. JSON — an array of records (first row as keys when marked as a header) for developers. XML — a clean structured document. HTML — a ready-to-embed web table. Markdown — a pipe table for READMEs, wikis and GitHub. You can also copy the table as tab-separated text to paste straight into a spreadsheet.
Because it rebuilds tables from the position of the text, detection focuses on data and grid structure. Preserved: cell values, row and column arrangement of clearly-spaced tables, Unicode/Hindi text, and multiple tables per page. Approximated or not fully automatic: merged cells (a spanning value lands in one cell), nested tables, borderless multi-column layouts, and cells whose text wraps onto several lines (joined with spaces). Original fonts, colours and borders from the PDF are not carried into the data formats; the Excel export applies a clean default style. That is exactly why the preview is editable — fix any imperfect split before exporting. OCR accuracy depends on scan quality.
“No table found on this page.” The page may be a scan — tick Use OCR and choose the language, then extract again; or the page simply has no tabular text. Columns look merged or split: the table is likely borderless or unevenly spaced — use the edit tools to add/delete a column or fix a cell. Numbers look odd or codes lost leading zeros: exported CSV keeps text as-is; in Excel, format the column as text if needed. Hindi text looks wrong: for text PDFs it is preserved as-is; for scans, choose the Hindi (or English + Hindi) OCR language. OCR is slow: it runs locally and the language model downloads once; large scans take longer — use Cancel to stop. “This PDF is encrypted.” Remove the password with our Unlock PDF tool first.