Because this piece was focused on software instead of hardware, the challenge was finding ways to make the AI stories feel visual without a device doing the heavy lifting. Cocreator in Paint was easily one of the hardest features to solve. The idea of sketching something rough and then letting AI turn it into an image sounds simple, but it was surprisingly hard to make readable on screen.

My first pass used a finished illustration, then layered in sketch marks and transitioned to the AI result. It was clear pretty quickly that it was not working. It explained the feature in theory, but it did not feel believable or intuitive.

After that, I spent more time in the product itself and rebuilt the sequence around a more accurate flow. I also tried a 3D version, but that pass ended up feeling too literal and too polished for the more fantastical tone the film wanted. After a few more rounds and some directorial feedback, I rebuilt the entire scene again and finally got it to a place where the feature made sense and still fit the larger video.

The Copilot section was one of the most complicated parts of the whole film. Early on, I built the story around step-by-step scenarios that showed what the product could do, but the first concepts stretched too long and started to feel heavy.

I narrowed those ideas down to three directions, find, recipe, and analyze, but even then the sequence still felt like it was asking the viewer to do too much. I also tried a voice-over track for a while, then moved away from it and let music do more of the work. That shift helped a lot because it forced the UI to be simpler and clearer.

Once the interaction design felt right, I moved the sequence into 3D and started building style frames around it. The final version was much more direct. It showed what Copilot could do without overexplaining it, and that simplicity ended up helping anchor the rest of the sizzle.