Split PDF by content
Split a single PDF into multiple PDFs using AI to detect logical sections.
Split PDF by content uses an AI model to detect logical sections within a single PDF, such as separate invoices, statements, or chapters bundled into one file, and splits it into one PDF per section. Unlike a page-based split, this transform reasons about the document's content to decide where each section starts and ends.
Availability: Requires the
WorkflowsPdfContentSplitfeature. Contact your OneSchema support representative if you do not see this transform.
What it does
- Input: one or more files (PDFs).
- Output: multiple PDF files per input file, one for each detected section.
- The transform analyzes the document's content — not just page breaks — to determine where one logical section ends and another begins.
When to use it
- A single PDF contains multiple invoices, statements, or records batched together by your vendor.
- Sections don't align to a fixed number of pages, so a simple page-count split won't work.
- You plan to run Extract PDF data afterward and want the model to work on one section at a time for better accuracy.
How to configure it
In the MFF builder, add a Split PDF by content node after the node that produces the PDF files.
Settings
| Setting | Description | Default |
|---|---|---|
| Prompt | A plain-language description of how to identify section boundaries, such as what marks the start of a new record. | Empty |
| Provider | The AI provider used to run the content analysis. Leave as the default unless your organization has a specific requirement. | Default provider |
Prompt tips
- Describe the boundary:
"Each invoice starts with a page that has an 'Invoice Number' header. Split into one PDF per invoice." - Call out exceptions:
"Treat the final page, which is a summary, as its own section."
Example
Input
A 12-page PDF containing four 3-page vendor statements batched into one file.
Output
Four separate PDF files, each containing one 3-page statement.
Troubleshooting
Sections are split in the wrong place
- Make the prompt more specific about what marks a new section, such as a recurring header or field.
Too many or too few sections are produced
- Add examples of what should and shouldn't count as a new section in the prompt.
- If sections are strictly page-based, use Split PDF by page instead and recombine downstream.
I just want one PDF per page
- Use Split PDF by page, which splits deterministically without AI.
See also
Updated 1 day ago