Convert Image to Excel

Turn a screenshot, scan or photo of a table into an editable Excel file. Check and fix cells in the browser, then download XLSX or CSV. No upload.

Your files are processed in your browser and never uploaded.

Numbers
Numbers with leading zeros, phone numbers and long ID numbers always stay as text, so Excel can't change them.

Drop images here

Choose files

HEIC, JPG, PNG, WebP, AVIF, GIF, BMP · up to 50 megapixels each · you can also paste

Works with screenshots, scans and photos of printed tables in English, Chinese and Japanese. Handwriting is not supported. The AI models (about 16 MB) download once, then stay in your browser's cache.

    How to use

    1. Drop one or more pictures of tables, or click "Choose files". Screenshots, scans and phone photos all work.
    2. The text is read and arranged into rows and columns. Cells that were hard to read are highlighted in yellow.
    3. Click any cell to correct it, then download the Excel file or CSV, or copy the table and paste it straight into a spreadsheet.

    Features

    Rows, columns and merged cells
    Ruling lines are traced to rebuild the grid, including merged cells and cells with several lines of text. Tables without lines are split into columns by how the text lines up.
    Check before you download
    The result appears as an editable grid next to your image. Unsure cells are marked, so you know exactly where to look instead of rereading the whole table.
    Numbers Excel won't mangle
    Plain numbers are stored as numbers so you can sum them. Codes with leading zeros, phone numbers and long ID numbers stay as text, so Excel can't round them or drop digits.
    Private, free and unlimited
    The AI models run in your browser. Your images are never uploaded, there is no sign-up and no page limit, and several images can be converted in one go.

    How the table is read

    Converting a picture into a spreadsheet takes three steps. First a text detector, PP-OCRv5 from the open-source PaddleOCR project, finds every line of text on the page. If the picture is slightly rotated, the page is turned upright by the average slant of those lines, which keeps rows straight. Each text line is then cut out and read by PP-OCRv6, a recognition model that knows around 18,000 characters. Finally, the words are placed into rows and columns, which is the part that turns plain text recognition into a usable table.

    Tables with and without lines

    Most tables have ruling lines, and those are the most reliable guide. The tool looks for thin dark lines running across and down the page, joins pieces of the same line that are interrupted by antialiasing or JPEG artefacts, and keeps the lines that cross each other to form a grid. A boundary that is missing between two neighbouring cells means they are merged, so a cell spanning two rows in your image becomes a merged cell in Excel. Coloured header rows count too: the edge of a filled area works like a line. Text that sits above or below the grid, such as a title or a note, is added as extra rows.

    When a table has no lines at all, or only a rule under the header, columns are found from the layout instead. Lines of text that share a height form a row. The tool then looks at which horizontal positions are covered by text in most rows; the gaps that stay empty all the way down are column boundaries. Text the detector read across two columns, which happens when the gap between them is narrow, is split at those boundaries.

    Getting the cleanest result

    Crop the picture to the table before converting it if the page also contains photos or other text, for example with our crop tool. For a long table on a website, take a full-page screenshot rather than a phone photo of the screen. If you only need the words and not the grid, use Image to Text, which keeps the reading order of paragraphs. To send the converted pictures together as one document, JPG to PDF combines them into a single file.

    Frequently asked questions

    What kind of images work best?

    Screenshots of tables from web pages, PDFs or apps give the best results, followed by flatbed scans and sharp phone photos taken straight on. Make sure the text is in focus and at least about 10 pixels high; zoom in before taking a screenshot of a small table. Photos taken at a strong angle, in dim light or with glare on glossy paper are harder, and handwriting is not supported.

    Which languages does it read?

    Printed English, simplified and traditional Chinese, Japanese, digits and common symbols such as currency signs, percentages and dates. A table can mix languages freely, for example Chinese column headers with English product codes.

    How accurate is it?

    On clean screenshots nearly every cell comes out right. Mistakes, when they happen, are usually look-alike characters (a lowercase l read as a capital I, O as 0) or a stray mark on a photo. Each cell gets a confidence score and the unsure ones are highlighted, so checking a result takes seconds. Always check important figures before you use them.

    Why are some numbers stored as text?

    Excel converts anything that looks like a number, which breaks codes. 007 becomes 7, a 13-digit order number is shown as 1.23E+12 and an 18-digit ID number loses its last digits. To avoid that, numbers with leading zeros, numbers with 11 or more digits and anything longer than Excel's 15-digit precision are kept as text. Choose "Keep as text" to store every cell exactly as written.

    Can I convert several images at once?

    Yes. Drop as many images as you like; they are read one after another, and each gets its own editable table. "Download all" saves one Excel workbook with one sheet per image, named after the files.

    Are my images uploaded?

    No. The page downloads the AI models (about 16 MB, once) and reads the images on your device. You can check this in your browser's network panel. No image ever leaves the page, so it is safe for invoices, payslips and other private documents.

    Related tools