How We Turned a 35-Question Quiz Into a 35-Slide Visual Deck

This project started with a question: what if every question on a quiz had its own picture? Doing that by hand would mean writing 35 image descriptions, generating 35 images, and building 35 slides. For perspective, presentation designer Nancy Duarte's book Slide:ology estimates 20 to 60 hours just to build the slides for a 30-slide presentation. This deck took me about five minutes of hands-on work, and the AI tools did the rest while I was teaching.

We did it with a relay of AI tools, where each one handled the part it does best.

Step 1: The quiz goes in. I loaded a finished 35-question quiz into Claude. Everything that followed was built from that one file, so the images line up exactly with what the quiz covers.

Step 2: Claude writes 35 image prompts. Claude wrote one detailed image prompt for each question. Every prompt used the same locked-in style description so the whole set would look like it came from one artist, and each one described the image as a single slide-sized scene. Claude saved all 35 prompts into one Google Doc in a shared Google Drive folder.

Step 3: A Grok bot runs the prompts. Instead of copying and pasting 35 prompts myself, I had a Grok bot do it. It took each prompt from the doc, ran it through Google's Nano Banana image tool in Gemini, and saved each finished image to the same Drive folder, numbered 01 to 35 with titles so they sort in slide order. I had it set up to switch over to Grok's own image tool if Gemini hit its usage limit, but this time Gemini finished all 35 without stopping.

SCREENSHOT GOES HERE

Step 4: Claude builds the deck. I handed the folder of images back to Claude, and it assembled them into a PowerPoint, one image per slide, in quiz order. Because each image was generated at slide size, they dropped in cleanly without cropping or resizing.

The work happened while I taught. My part was loading the quiz, starting the bot, and handing the images back at the end. Everything in between ran on its own during my class periods. By the time I had a free moment, the images were waiting in the folder.

Why the relay works. No single tool did everything well, so we didn't ask it to. Claude is strong at writing detailed, consistent prompts and building files. The image tools are strong at drawing. The Grok bot handled the most boring part, which was feeding prompts in and saving images out one at a time. Keeping everything in one Drive folder meant each tool knew exactly where to find its input and where to put its output.

What to watch for. Image tools have daily limits, so a 35-image job can stall halfway. Having a second image tool ready as a backup means the job keeps moving if that happens. It's also worth reviewing every image before using the deck. When the bot finished, it flagged one image, a Separation of Powers slide, that drifted from its prompt. The prompt called for one big lightning bolt splitting into three, lined up with three government buildings, and the image showed several small bolts that didn't line up. One image out of 35 needing a redo is a good result, but you only catch it if you look.

The same relay works for any set of materials that follows a list, whether that's vocabulary terms, review questions, or the steps of a process. Start with the list, turn each item into a prompt, let a bot run the prompts, and build the deck at the end.