https://www.instagram.com/p/Dc4DBdSM30w/
Hey, this is Yutao. Welcome to the tutorial.
Here's exactly how I turned a normal clip of me at my desk into that shot where I get pulled into a Pac-Man maze.
Real talk: my first version took hours of trial and error to dial in. But that's the whole point of this guide. I already figured out the prompts and the workflow, so you can skip straight to the part that works.
🟢 The only tool you need for this is Higgsfield. It's basically a one-stop shop. Every major AI model in one place, so you're always on the newest ones. 👇
Higgsfield AI — AI-native creative suite
That's it. Budget an afternoon your first time. Way faster once you know the flow.
1a. Film your raw shots. Shoot yourself at your desk from a few angles so you have options to work with later:

.png)


1b. Generate your reference image. Take a frame from your footage into Higgsfield and overlay the game scene onto your monitor. This becomes the "control" image that keeps everything consistent in the next step. Here are some reference prompts that you can use
<aside>
Replace only the green screen area of the monitor with a classic Pac-Man arcade maze — black background, neon blue maze walls, yellow dots, yellow Pac-Man mid-chomp, colorful ghosts, retro arcade style with score text in the corner. Important: the monitor is angled away from the camera, so its right side is farther away than its left. Display the maze at that same three-quarter perspective — left edge closer and taller, right edge receding and shorter — so it conforms to the tilted screen like a real display, not a flat poster. Keep the person, room, desk, and everything else exactly the same.
</aside>



