cappv1 reads a PowerPoint deck or Word document, asks an AI model to describe each picture, chart and table, and writes the answer into that item's alt text. Nothing else in the file changes.
The only thing cappv1 ever writes is a shape's alt text, the same field PowerPoint's own Edit Alt Text panel changes. Layout, positions, formatting, pictures, text, notes and slide order are left exactly as they were. It also marks invisible, empty shapes as decorative, which is a pure label and changes nothing you can see.
image1.png counts as no alt text, because some tools fill that in automatically.AI descriptions are good starting points, not guarantees. They can miss context or get details wrong. Open the finished deck, check the alt text in PowerPoint, and edit anything that is off, particularly for charts with exact numbers and for public or legally required material.
cappv1 can be used from the terminal (type cappv1) or from this web page (/web in the terminal, or cappv1 --web). Both do the same work, share the same settings, and see the same run.