PDF 转 CSV - 将 PDF 中的所有表格提取为 CSV

上传 PDF，工具会检测每一页中的每个表格，并将每个表格导出为各自独立的 UTF-8 CSV 文件。如果 PDF 含有多个表格，您将得到一个 CSV 的 ZIP 压缩包——每个表格一个文件——这样您就能直接将它们导入各自的工作表、数据库或 BI 工具，无需手动切分。

每个识别到的表格都会生成各自的 CSV，并打包成 ZIP 下载——例如 bank-statement-q1.csv、bank-statement-q2.csv 等等。表头相同、跨多页的表格会自动拼接为单个 CSV，因此续表行不会产生重复文件。

可以。该 CSV 输出符合 RFC-4180 标准，采用 UTF-8 编码和 CRLF 行结束符，所有现代电子表格（Excel、Google Sheets、Numbers、Airtable）和 BI 工具（Tableau、Power BI、Looker、Metabase）都能在没有配置提示的情况下导入。拖入文件即可立即开始分析。

单层合并表头会被智能拆平——合并单元格的值会在其跨越的各列中重复填充，使每一行都拥有完整的列。子表头行会被自动识别，并根据你在下载前的偏好设置，要么作为前缀内联，要么保留为单独的一行。

可以。货币符号（$、€、£、¥、₹）、日期格式（任何区域）、百分比以及 Unicode 字符（带重音的字母、中文、阿拉伯文、表情符号）都能在转换中完好保留。引号和转义遵循 CSV 规范，因此导入 Postgres、MySQL 或 SQLite 时能忠实保留数据。

上传的文件会在几分钟内删除，绝不用于训练，绝不共享。CSV 输出没有水印、没有署名列、也没有跟踪像素——它是完全归你所有的纯数据。你可以将它用于 ETL 流程、CRM 导入、销售管道分析以及任何商业工作流，无需担心授权问题。

Yes — every detected table opens in an editable grid before download, so you can click any cell to correct an OCR misread, add a missing row with the + button, or delete a row that shouldn't be there. Your edits are what gets written into the final CSV, not the raw extraction.

琥珀色标记出 OCR 引擎置信度不高于 80% 的单元格，让你在信赖导出结果之前，清楚知道该重点核对哪些数字或文字。编辑被标记的单元格会清除其高亮，未改动的单元格则会保留警示，直到你完成核对。

Yes — click Reset to discard every manual edit and restore the table exactly as it was first extracted from the PDF, without re-uploading the file. This is the fastest way to back out of a correction pass that went wrong.

The editable verify grid only appears when a PDF produces a single table; as soon as more than one table is detected, the tool packages them straight into a ZIP of separate CSVs with no inline review step, since there is no single grid left to show. If you need to check numbers before trusting them, open each CSV in a spreadsheet first, or use PDF Split to isolate one table's pages and re-run just that page range through PDF to CSV to get the verify-grid treatment.

You can convert PDFs to CSV without creating an account. If you sign in on a free plan, a small usage meter appears above the upload box once you're close to that day's free extraction allowance, since table extraction runs OCR behind the scenes — Premium accounts don't see this limit at all.

Scan the grid row by row for the classic OCR slips — a merged column, a dropped decimal, a stray header line picked up as data — since those are the mistakes most likely to throw off a reconciliation. Use Add row to restore a line the detector merged or skipped, Delete row to drop stray header or footer text read in as a row, and Reset if a correction pass goes wrong and you'd rather start over from the original extraction. Only download once the grid matches the statement, so the CSV that reaches QuickBooks or Xero is already clean.

No — there is no such preference control anywhere in the tool; sub-header detection and placement is fully automatic and cannot be switched by the user, despite what some site copy implies. If the automatic result puts a sub-header in the wrong place, use the editable verify grid (shown for single-table results) to fix or delete that row with the cell editor, Add row or Delete row before downloading, or adjust the CSV afterward in a spreadsheet.