Can table image to CSV preserve merged cells?

CSV has no merged-cell feature. Choose a flat set of columns and review how spanning headings should be represented in your records.

What if two tables appear in the same image?

Crop and process each table separately when their columns differ. This avoids confusing unrelated headings or mixing incompatible records.

How do I distinguish a blank field from a zero?

Compare the image and review grid carefully. Missing content and the number zero can have different meanings and should not be substituted automatically.

Should I choose CSV or searchable PDF for a table image?

Choose CSV for field-based import and analysis. Choose searchable PDF to retain the visual table and search recognized words within the document.

Why is the last column of my CSV mostly empty?

Rows are padded to the widest row, and the grid has no column controls, so one wide row leaves empty cells in every other row. Delete the empty column in a spreadsheet after opening the CSV, or ignore it in the importer's column mapping.

Does the CSV include the title printed above the table?

A title printed above the table arrives as a single-cell row at the top of the grid, and the first row of the grid is what becomes the header. Delete the title row so your column names come first, or the importer will treat the title as the header.

What does Table AI (beta) change for a difficult table?

It is an alternative table-recognition engine that suits tricky layouts with merged headers. Choose it in Settings and convert again, since changing the engine clears the grid. Default handles most tables, and Engine 1 and Engine 2 work best on simple grids.

How can I check that no rows were lost?

Count the data rows in the picture and in the grid, and compare the last row of each. If the table has a total or a numbered first column, use it as a check, since numbers should run without gaps. Use Add row for a missed line and keep the header row first.

Do columns keep the order they have in the picture?

Columns follow their left-to-right position in the image, and the grid cannot reorder them. If your destination needs a different order, map the columns in the importer or rearrange them in a spreadsheet after opening the CSV.

Can I convert a folder of table images at once?

Batch mode, a Premium option, converts many images with the same settings and returns a ZIP with one CSV per image. There is no review grid in batch, so use single-file mode for tables where a blank column or a shifted value would matter.

Image to CSV

Table Image to CSV — Extract Structured Text Rows

Choose your image

Choose one table image with all relevant columns visible

Turn an image of a table into CSV records. Review column boundaries, blank cells and repeated headings before saving data for a spreadsheet or script.

Reference-table ingestion
Borderless table cleanup

Upload your table image

Choose one table image with all relevant columns visible

All image formats (JPG, PNG, WEBP, HEIC, TIFF…)

Table image to CSV: consistent fields from a table trapped in a picture

This table image to CSV page treats any picture of a table, scanned, photographed or exported, as a source of records. It reads the cells into a grid, you make sure every row has the same fields, and you save a CSV that a script or database can take. A generic table has no fixed meaning, unlike a receipt or a statement, so the checks are about structure: column count, headers, blanks and types.

Make every row the same width

A CSV is only valid data when each row has the same number of fields. The converter pads short rows with empty cells at the end, so a value that slid left leaves a blank at the right edge instead of where it belongs. Scroll the grid and look at that right edge: a column that is empty except for a few rows is a sign that something moved. Retype the value under the right heading.

Flatten two-level headings into one row of names

Tables often print a group heading above two or three sub-columns, such as Q1 above Sales and Units. A CSV has one header row, so combine the levels into names like Q1 Sales and Q1 Units by typing over the header cells, and delete the row that held the group name. Do this before you export, because the first row of the grid is what the file uses as its header.

Tell a blank from a zero and from a dash

Printed tables use a blank, a dash, n/a and 0 for different reasons, and the grid shows whatever was read. Decide a convention for the destination, such as leaving missing values empty and writing 0 only where 0 was printed, and apply it consistently before export. A dash can be a minus sign or a missing value, which only the context of the table settles.

Crop the picture to one table and its header rows.
Convert the file and check every row for width.
Flatten the headers, then settle blanks, dashes and zeros.
Choose See result and Download, then test a small import.

Keep notes, units and footnotes out of the data

A table image often includes a title, units in a caption and footnotes. They arrive as single-cell rows, and two tables in one picture are stacked with a blank row between them. Write units into the column names first, such as Weight kg, because a CSV cannot carry a caption, then delete the rows that are not records. Footnote markers inside cells stay part of the value until you remove them.

Before you export the table CSV

Keep one table per file, and crop a second table into its own image.
Choose comma and UTF-8 in the importer before you test a small load.
For a spreadsheet you can format, convert the same table to Excel.
For a readable table in a README, use Markdown instead.

Get consistent fields out of a table trapped in pixels

Convert a visible table into rows of text that your tools can import. The useful result is a consistent set of fields, so inspect column boundaries and missing values before treating the export as structured data.

Reference-table ingestion

Prepare a reviewed CSV from an image-only reference, then validate field types and allowed values before a script consumes it.

Borderless table cleanup

Check whitespace-separated columns in the editable grid and restore blank fields that would otherwise shift a row's meaning.

Check table boundaries and blanks before exporting records

Isolate a single table while retaining the headings needed to identify fields.
Inspect rows with blank cells, wrapped labels or spanning headers.
Export CSV and validate record width and data types in the destination.

Merged headings, multiple tables and blank-versus-zero values

Can table image to CSV preserve merged cells?: CSV has no merged-cell feature. Choose a flat set of columns and review how spanning headings should be represented in your records.
What if two tables appear in the same image?: Crop and process each table separately when their columns differ. This avoids confusing unrelated headings or mixing incompatible records.
How do I distinguish a blank field from a zero?: Compare the image and review grid carefully. Missing content and the number zero can have different meanings and should not be substituted automatically.
Should I choose CSV or searchable PDF for a table image?: Choose CSV for field-based import and analysis. Choose searchable PDF to retain the visual table and search recognized words within the document.
Why is the last column of my CSV mostly empty?: Rows are padded to the widest row, and the grid has no column controls, so one wide row leaves empty cells in every other row. Delete the empty column in a spreadsheet after opening the CSV, or ignore it in the importer's column mapping.
Does the CSV include the title printed above the table?: A title printed above the table arrives as a single-cell row at the top of the grid, and the first row of the grid is what becomes the header. Delete the title row so your column names come first, or the importer will treat the title as the header.
What does Table AI (beta) change for a difficult table?: It is an alternative table-recognition engine that suits tricky layouts with merged headers. Choose it in Settings and convert again, since changing the engine clears the grid. Default handles most tables, and Engine 1 and Engine 2 work best on simple grids.
How can I check that no rows were lost?: Count the data rows in the picture and in the grid, and compare the last row of each. If the table has a total or a numbered first column, use it as a check, since numbers should run without gaps. Use Add row for a missed line and keep the header row first.
Do columns keep the order they have in the picture?: Columns follow their left-to-right position in the image, and the grid cannot reorder them. If your destination needs a different order, map the columns in the importer or rearrange them in a spreadsheet after opening the CSV.
Can I convert a folder of table images at once?: Batch mode, a Premium option, converts many images with the same settings and returns a ZIP with one CSV per image. There is no review grid in batch, so use single-file mode for tables where a blank column or a shifted value would matter.