Document Preparation

The Document Preparation tool lets you preprocess PDFs and images before uploading them. Split multipage documents, resize images, adjust contrast, remove duplicates, and more — all through reusable pipelines you can save as recipes.

Purpose

When working with scanned documents, raw images are rarely ready for direct use. Pages may be tilted, oversized, or bundled into a single PDF that needs splitting. Document Preparation solves this by providing a pipeline of image operations you chain together and run in one click.

Standardize scanned pages to consistent sizes and orientations
Clean up images with deskew, contrast adjustment, and grayscale conversion
Split multipage PDFs into individual images
Pack multiple pages into composite images for review
Remove unwanted pages or duplicate scans
Mask sensitive regions before sharing

Key Concepts

Recipes

A recipe is a saved pipeline of processing steps. Create it once, then reuse it whenever you have similar documents. Each recipe has a name, optional description, and an ordered list of steps.

Steps

Each step performs one operation (e.g., resize, rotate, crop). Steps run in order — the output of one step becomes the input of the next. You can enable or disable individual steps without removing them.

Runs

A run is a single execution of a pipeline. It tracks the input files, output files, execution time, status, and a detailed log of each step. Runs can be linked to a saved recipe or be ad-hoc (Quick Process).

Quick Process (One-Time)

Navigate to Doc Prep in the sidebar
Click Start on the Quick Process card
Add processing steps using Add Step
Upload PDF or image files via drag-and-drop or the file browser
Click Execute to process
Download the results when complete

Quick Process is ideal for one-off tasks where you don't need to save the pipeline for later.

Creating a Recipe

Navigate to Doc Prep → New Recipe
Enter a name and description
Build the pipeline by adding steps in order
Configure each step's parameters (e.g., DPI for PDF conversion, max dimensions for resize)
Click Save Recipe

You can also clone an existing recipe to use it as a starting point.

Running a Recipe

Open the recipe's detail page
Upload files in the Execute Pipeline section
Optionally enter a run name
Click Execute
Monitor progress in real time
Download output files when complete

Recent Runs

The main Document Preparation page shows all your recent pipeline runs across recipes and ad-hoc executions. Each entry displays the run name, the recipe name (if linked), status, timing, and a download link.

Search

Use the search box to find runs by run name or recipe name. The search is case-insensitive and matches partial text.

Date Filters

Click the Date dropdown to narrow results by time range:

Last hour — runs from the past 60 minutes
Today — runs since midnight
Last week — runs from the past 7 days
Custom range — pick specific start and end dates

Pagination

Runs are displayed 5 per page with Previous/Next navigation. All filters are preserved when navigating between pages.

Supported Operations

Step	Description
PDF to PNG	Convert PDF pages to individual PNG images at a configurable DPI (72–600). Maximum 200 pages per PDF.
Resize	Scale images to fit within maximum width and height, with configurable quality (1–100%)
Pack	Combine multiple pages into single composite images (vertical or horizontal). Configurable pages per group (2–20) and optional gap between pages.
Stitch	Merge overlapping frames (panorama capture of a long or wide document) into one single image. Alignment is computed from the image content; frames without enough overlap are joined edge-to-edge along the configured pan direction (vertical or horizontal).
Rotate	Rotate images by 90, 180, or 270 degrees
Crop	Two modes: trim margins (by pixel amount per edge) or extract a specific region (absolute coordinates)
Split	Split images along one or more lines. Supports horizontal and vertical splits in a single step — horizontal splits are applied first, then vertical.
Grayscale	Convert color images to grayscale
Deskew	Straighten tilted scanned pages (configurable threshold)
Contrast	Adjust brightness and contrast levels independently
Remove Pages	Drop specific pages by number
Reorder Pages	Rearrange page sequence
Deduplicate	Remove near-duplicate pages using perceptual hashing. Images are compared via a 16×16 grayscale fingerprint with configurable similarity threshold (50–100%, default 95%).
Mask Region	White-fill rectangular areas to hide sensitive content

Use Cases

Preparing Scanned Invoices for Text Extraction

Create a recipe with: PDF to PNG (300 DPI) → Deskew → Contrast (increase) → Resize (max 2000px). This produces clean, straight, consistently-sized images ready for accurate text extraction.

Splitting a Multi-Page Contract

Use PDF to PNG to convert a 20-page contract, then Remove Pages to drop cover pages and blank pages. The output is a clean set of content pages ready for form instance creation.

Redacting Sensitive Information

Apply Mask Region to white-fill areas containing personal data (SSN, account numbers) before sharing documents externally. Chain with Grayscale for a consistent look.

Creating Review Sheets

Use PDF to PNG → Resize (small thumbnails) → Pack (4 per group, vertical) to create compact review sheets from multipage documents. Ideal for quick visual comparison.

Cleaning Up Phone Photos

Chain Deskew → Crop (remove borders) → Contrast → Resize to turn rough phone photos of documents into clean, standardized images suitable for archiving or automation.

Deduplicating Bulk Scans

When scanning large batches, duplicate pages are common. Use Deduplicate (threshold 85%) to automatically remove near-identical pages, then Reorder Pages to arrange the remaining pages correctly.

Managing Recipes

Edit — Modify steps, reorder, enable/disable individual steps
Archive — Soft-delete a recipe (removes it from the list)
Clone — Create a copy of an existing recipe as a starting point

Execution Log

Each run records a detailed per-step execution log. After a run completes (or fails), you can see:

Step type and position — which operation ran and in what order
Status — completed or failed
Duration — how long the step took (in milliseconds)
Input/output counts — how many files went in and came out
Error message — details if a step failed

When a run produces multiple output files, they are automatically bundled as a ZIP archive for download.

Tips

Steps run in order. Place PDF to PNG first when working with PDFs.
Disable steps temporarily instead of removing them — you can re-enable them later.
Use Quick Process to experiment before saving a recipe.
The search in Recent Runs matches both the run name and the recipe name, so you can find all runs for a specific recipe quickly.
Accepted file types: PDF, PNG, and JPG/JPEG.