Technical Requirements & Platform Support
Will PAC be available on macOS?
Yes, a macOS version is already in development. A first prototype will be presented at the axes4 Day in March 2026, one of the key industry events for document accessibility
Will other axes products be available on macOS as well?
We are working on a macOS version of axesPDF, which is scheduled for release in the second half of 2026.
What are the hardware requirements to run the model locally?
The AI model is small and has no special requirements. A standard PC/laptop is sufficient for processing.
Does PAC run completely offline?
The AI model is delivered with PAC 2026 and runs entirely locally. PAC does not access the internet in any way.
Is PAC GUI-only, or are there plans for a command-line interface?
No. We provide axesSense for this purpose. axesSense is available as a web service or, upon agreement, as a local installation.
Is CPU/AI power (e.g., NPU) utilized when processing on a laptop/desktop?
NPUs (Neural Processing Units) are not used. However, if technically possible, PAC-AI uses the GPU (graphics card) for faster processing.
Functionality & Integration
Are you planning to release an add-on (similar to PAC 2026 AI) that highlights/fixes accessibility issues directly within Microsoft Word or other tools? (Advanced Accessibility Check)
We have no plans for an additional add-in with such functionality. However, we are considering expanding axesWord accordingly.
What software do companies use to ensure documents are truly and semantically accessible?
For creating fully accessible documents, we recommend axesWord, axesSlide, or axesFlip. To check them, we recommend PAC or axesSense.
We recommend axesPDF for editing PDF documents with regard to accessibility.
How can we ensure a PDF is tagged in a way that produces accurate output for screen readers, especially to avoid duplicated content when both regular content and alternative texts are present?
Our general recommendation is to tag documents in accordance with the applicable standards (Matterhorn Protocol) and not for specific screen readers. We recommend the above-mentioned programs from axes4 for this purpose.
If documents are correctly tagged in accordance with the applicable standards, it is up to the screen reader manufacturers to ensure correct output.
Standards, Evaluation & Feedback
What type of data was used to train the AI model?
axes4 has a set of high-quality accessible PDF documents. We used this set to train the AI model.
If we notice incorrect AI judgments, what specific information would be helpful for reporting and how should we submit it?
If possible, please send the relevant PDF document to support@axes4.com with an explanation of the error.
In the U.S., there are many agency-specific human checklists. Is there a more centralized or standardized approach in Europe for the human component of accessibility evaluation?
The internationally applicable Matterhorn Protocol contains a column entitled "How," which defines whether a check can/must be performed by a machine or by a human. We recommend this document as a point of reference.
For Germany, there are recommendations from the BIT-inklusiv network on testing procedures for PDF documents.
Please sign in to leave a comment.