Deck Rescue FAQ
What is Deck Rescue?
A web page that fixes two problems Ally flags in image-heavy PowerPoint decks: missing slide titles and missing image descriptions (alt text). You open a .pptx, see every slide on one page, fill in what's missing, and download a fixed copy to upload to Canvas.
What does the AI do?
For each picture that needs a description, you can click Suggest. An AI model drafts a description, and you accept, edit or discard it. Nothing the AI writes goes into your file unless you accept it.
What doesn't it fix?
Charts, tables, SmartArt, color contrast and reading order. It points out charts, tables and SmartArt so you can fix them in PowerPoint. It works on .pptx files only, not Google Slides, Keynote or PDF.
Does it change my original file?
No. You download a zip ending in "-rescue.zip". Unzip it, then open START HERE.html. Inside are a fixed copy with the same name as your original, so it can replace the original in Canvas, and a short report of what was changed and what still needs work. The file on your computer isn't touched.
Where does my file go?
Your file stays on your computer; the page does its work in your browser. When you click Suggest, that one picture, shrunk to a smaller size, plus the text and speaker notes on its slide, are sent through this site's server to Google's Gemini service to draft a description. Google doesn't use it for training and keeps it for 55 days to check for abuse. Deck Rescue doesn't store pictures or descriptions.
Can I use it on slides with student work?
Deck Rescue can't tell what's in a picture or decide whether it's appropriate to send. Whether to use Suggest on slides showing students or student work is each instructor's decision, made with RVCC's AI Usage Policy in mind. That policy, for example, lets a student ask that their work not be entered into AI tools, and the instructor must honor the request. You can always type descriptions yourself.
Does it raise Ally scores?
Decks fixed with its method scored 100% on titles and alt text in an Ally sandbox course. Your course score depends on everything else in it too.
How good are the AI descriptions?
They are first drafts. The AI is told to describe what a sighted student would get from the picture, following museum guidelines for describing art. Read each one before accepting it.
Because Deck Rescue is being developed for RVCC faculty, the instructions given to the AI can be refined as faculty needs come up. That's a real advantage over the AI-generated alt text built into programs like PowerPoint, whose instructions no one at RVCC can see or change.
Is Deck Rescue approved by the college?
RVCC hasn't yet published a list of approved AI tools; the college's AI Steering Committee is developing that guidance. AIDL has encouraged sharing the version of Deck Rescue without AI, and I've described the AI side to AIDL and asked for its guidance. Deck Rescue's AI drafting only works with an access code, and I issue each one personally to an RVCC email address, so it isn't used without my knowledge, and I can turn any code off.
AI-drafted alt text is already part of tools faculty have through the college, such as PowerPoint's automatic alt text and the DesignPLUS Transformation Wizard in Canvas. RVCC's AI Usage Policy leaves AI use in teaching and course materials to each instructor, within its data rules. Deck Rescue is built with those rules in mind: Google's paid service doesn't train on what's sent, and nothing the AI writes reaches your file until you accept it.
Who pays for it?
The AI drafts cost money per picture. Right now the liaison for Arts and Design covers that cost personally, so each user gets an access code and a daily limit.
How do I get access?
Email Cynthia Teeters (Cynthia.Teeters@raritanval.edu) for an access code and instructions.
Is the page itself accessible?
Yes, and that was one of the reasons for building it. I have low vision, and I found PowerPoint's own Accessibility Check panel unfriendly and itself lacking in accessibility: I couldn't enlarge its font size.
Deck Rescue sizes all its text so browser zoom and your font settings enlarge everything. It scores 100 in Lighthouse's accessibility check, works by keyboard alone, and reflows at 400% zoom without sideways scrolling.
It hasn't yet been tested with a screen reader. That's a particular gap for me, since I haven't been able to configure and use VoiceOver, the screen reader on my Mac, well enough to test with it. I hope to resolve that, or to get feedback from someone who uses a screen reader every day. If that's you, I'd love to hear from you.
Which browsers work?
Tested in Chrome. Safari and Edge haven't been checked yet.
Who maintains it?
Cynthia Teeters, Digital Accessibility Liaison for Arts and Design, who wrote it and can change it when faculty need something different.