PDF Extract FAQ

No. While standard OCR only turns images into text, Supabind uses Large Language Model (LLM) Vision. This means the AI actually understands the document layout, identifying where headers, rows, and nested tables exist, even if they aren't explicitly marked.

Gemini 2.5 Flash (Balanced) is optimized for speed and works best for standard digital PDFs. Gemini 2.5 Pro (High Accuracy) has deeper reasoning capabilities, making it the choice for low-resolution scans, handwritten notes, or extremely complex catalog layouts.

It represents the AI's internal probability that the extracted table structure and text content match the source document. A score above 90% is production-ready; scores below 70% usually indicate that the PDF was too blurry or the layout was too ambiguous.

Yes. The AI identifies all tabular data on a page. In the Extraction Results editor, you can use the 'Table Selector' dropdown at the top of the grid to switch between different tables found in the same document.

Our Gemini 2.5 models have a large 'context window,' allowing them to see the whole document at once. If you use the Balanced or High Accuracy mode, the AI will automatically stitch rows across page breaks into a single continuous table.

Extraction Prompts allow you to give the AI specific instructions. For example, if you only need columns for 'SKU' and 'Price', you can update the prompt to say: 'Extract only SKU and Price columns.' This significantly improves accuracy on dense catalogs.

The most common reasons are: 1) The file exceeded the 10MB limit. 2) The PDF is password-protected. 3) The document contains no readable text or tables. Check your 'Processing History' logs for the specific error code.

Yes. Because we use advanced LLMs, the extraction engine supports over 50 languages, including complex scripts like Arabic, Chinese, and Japanese, as well as multi-lingual documents.

Your files are processed in a secure, isolated environment. We use enterprise-grade APIs where your data is not used to train the underlying models. Files are encrypted at rest and can be deleted permanently from your 'My Files' tab at any time.

Currently, we support .pdf files. Support for image formats (.jpg, .png) and multi-page TIFF files is currently in development and will be released in a future update.

Yes. You can connect an FTP/SFTP server to Supabind. Once connected, you can set the AI to automatically process any new PDFs that arrive in a specific folder without manual intervention.