Learn AI-Powered PDF Auto-Tagging with PDFix Desktop

You can find all our past and upcoming webinars in the Webinar category on pdfix.net. More recorded sessions are also available on our YouTube channel: Team PDFix

Absolutely! You can follow along and experiment using the same materials. All the documents and JSON template files are available for download on our GitHub: Weekly Market Sample

Yep, everything you see in this webinar is replicable in the SDK. See a separate webinar for auto-tagging with PDFix SDK.

Templates can be created with the Preflight function in PDFix Desktop, using an AI model, or manually. Template examples are available on GitHub.

Yes, it’s possible. If you need help creating one, just contact us — we can assist with multi-page and repeating layouts.

Yes, that’s correct. Templates can be generated automatically using AI. PDFix Desktop then applies them to fix and enhance the tagging structure.

Some AI models require credentials, some don’t. PDFix Desktop offers a free AI layout models such as Paddle.

PDFix Desktop offers a free AI layout models such as Paddle which does not require credentials.

Yes. Each model has its own credentials or is free, depending on licensing. If you have your own, we can integrate it into the PDFix pipeline.

Yep! You can train and prepare your own model, then integrate it easily into PDFix.

We continuously work on imtegration of new LLM models into PDFix Desktop, including Microsoft Azure. Please check for PDFix Marketplace updates.

Yes! We offer an external OCR action that can be applied before auto-tagging. It automatically adds an OCR text layer to scanned PDF files. Learn more here: OCR Tesseract Action

Yes, we have an external Table Summary mode that improves table readability for LLMs: Generate Table Summary with OpenAI

We support both OpenAI and Paddle for MathML detection.

Tagging PDF Forms is challenging, but possible. Each form field is properly tagged based on PDF/UA standard. The form field descriptions can be auto-generated from field names or tooltips.

Manually created templates can handle auto-tagging of complex layouts. If you need help creating one, just contact us — we can assist with complex layouts.

Not completely. PDFix relies on the open-source veraPDF tool for PDF/UA validation. You can learn more here: PDF Accessibility Validators

Yep! Check our related blog and webinar here: How to automate fixes in InDesign created PDFs

Yes, exactly! You can use the free BLIP or paid OpenAI model for automatic Alternate Text generation.

Constantly. External actions are updated whenever new versions are released. PDFix SDK and Desktop are updated at least quarterly — or more often if needed.

Yes, it is possible to use Podman instead of Docker Desktop for running external actions integrated with PDFix Desktop.
PDFix Desktop supports actions that can be distributed through the PDFix Marketplace or installed manually using an action configuration file (JSON).
Each action configuration file defines:

• Action metadata — such as name, category, and subtype, enabling integration into specific workflows (e.g., template creation or tag editing).
• Program execution pattern — including the command-line call and its arguments.
• Argument definitions — allowing customization of how the action runs and interacts with PDFix Desktop.

Because the action system executes external programs through command-line calls, Podman can be used in the same way as Docker Desktop or any other CLI-based container runtime.
Example: Executing the Action with Docker

docker run -v $(pwd):/data -w /data --rm pdfix/autotag-textract:latest \
  tag --aws_id ${AWS_ID} --aws_secret ${AWS_SECRET} --aws_region ${AWS_REGION} \
  -i /data/input.pdf -o /data/output.pdf

Example: Executing the Same Action with Podman

podman run -v $(pwd):/data -w /data --rm pdfix/autotag-textract:latest \
  tag --aws_id ${AWS_ID} --aws_secret ${AWS_SECRET} --aws_region ${AWS_REGION} \
  -i /data/input.pdf -o /data/output.pdf

Since Podman provides a Docker-compatible command-line interface, no additional configuration changes are required in the PDFix action definition.
Simply replace docker with podman in the execution command. For detailed guidance on creating or installing custom actions, please contact us.

• Create or sign in to your AWS account (console).
• Create an IAM user (or use an existing one) and enable Programmatic access so it can get access keys. When you create the access key pair, save the secret — AWS shows the secret only once. AWS Documentation
• Attach Textract permissions to that user/role. For testing you can attach the managed policy AmazonTextractFullAccess; for production prefer least-privilege (grant only the textract:* actions you need). AWS Documentation
• Create the access key for that IAM user (Access Key ID + Secret Access Key) via the IAM → Users → Security credentials → Create access key UI. Store those credentials securely (see storage below). AWS Documentation

Paddle is currently supported only for layout recognition in auto-tagging and template workflows. It does not include OCR functionality.
At the moment, Tesseract is the only OCR engine supported in the PDFix Marketplace. Please check for future updates that may add OCR support for Paddle or other engines.