OCR may return every number while associating a price with the wrong product. Spellchecking alone will not catch that. Verify which item each value belongs to before reusing the data.
Characters and table structure
Tesseract documents limitations with table recognition without additional segmentation or layout analysis. FileMyLove's image to text provides text to edit, copy or download as TXT. It does not create spreadsheet cells from the image.
Spaces that resemble columns in the result do not prove correct placement, especially where descriptions wrap or source cells are empty.
Review rows with their headings
We recommend keeping the source image beside the text and starting with one easy-to-check row.
| Check | Question to resolve |
|---|---|
| Item label | Which item does this group describe? |
| Column heading | Is this quantity, unit price or total? |
| Symbols | Are decimal points, minus signs and units present? |
| Empty positions | Is the source empty, or was something unreadable? |
| Wrapped descriptions | Is this a new row or a continuation? |
Do not automatically replace an empty position with zero. A plausible total does not prove every entry is right. Mark unclear values for review instead of guessing from neighbouring rows.
Prepare a manageable section
- Retain the complete original image as a reference.
- For a long table, prepare a short copy that keeps the necessary headings and item labels.
- Leave space around characters rather than clipping digits at the boundary.
- Select the appropriate language and recognise that copy.
- Compare each row with the image, then place verified entries into a new table.
Splitting the image is a review technique, not an accuracy guarantee. Removing labels and retaining only numbers makes their relationships harder to check.
Prefer source data when available
If the sender can provide Excel or CSV, we suggest starting there instead of reading an image. Still check identifiers, values and headings after opening it.
For a scanned PDF, PDF OCR can provide searchable text but does not establish correct spreadsheet cells. See the general image-text guide for the basic workflow.
Open the downloaded TXT after correcting it and retain the image until review is complete. Processing runs in your browser without uploading the picture.
