Fly Dive

• 

  Scene 2

Inspiration

So basically, I've always been obsessed with skydiving videos on YouTube, but let's be real—who has $3000 lying around to jump out of a plane ? Plus, I'm honestly terrified of heights. During our brainstorming session, someone mentioned how cool it would be to experience that view without the whole "falling to your potential death" thing. That got me thinking about using AI to create realistic first-person videos. Then I extended it and added the sea diving part. But, my primary focus was to generate a hyper-realistic sky diving video in 360 degrees ultra HD quality video.

What it does

The man abandons his parachute and plunges into the vast sea below. As he dives deeper beneath the surface, he begins to explore the underwater world around him. Suddenly, a school of sharks encircles him, their sleek bodies gliding through the murky depths. Amidst the danger, he spots a first-aid kit resting on the ocean floor, along with an ornate mirror framed in gold. After navigating through the perilous waters, he finally swims toward the shoreline and emerges from the ocean, collapsing onto the sandy beach.

How I built it

Not gonna lie, I spent way too much time messing with AI video tools at 2 AM. I started by researching how actual skydiving videos look—camera angles, body positioning, lighting conditions. Then I built these really detailed text prompts that describe every single aspect of the scene. It took forever to get the hands to look right because they kept coming out super weird. I basically iterated on my prompts like 50 times until the output started looking realistic enough to fool someone.

Challenges faced

Dude, making a first-person perspective that doesn't look janky is surprisingly difficult. The AI kept generating these alien-looking hands with weird proportions. Also, getting everything to work together—clouds, ocean, sunset, parachute physics—was a nightmare. Sometimes the lighting would be perfect but the ocean looked fake, or the hands were good but the clouds were blocky. I probably watched the test videos a hundred times trying to spot what felt off. Sleep deprivation definitely became a thing toward the end.

Accomplishments

Honestly? The fact that it actually works blows my mind. When I finally got a video that looked convincing, I just sat there watching it on repeat. The hands look human, the sunset is gorgeous, and the whole thing feels way more real than I expected it to. I basically figured out how to give people a skydiving experience without them needing to leave their room or risk their lives. That's pretty neat.

What I learned

Prompt engineering is way harder than it sounds—every word matters. I also learned a ton about cinematography, like how camera angles and lighting completely change how realistic something looks. Working under a tight deadline taught me to prioritize features and not get stuck perfecting small details. Plus, I got better at debugging and staying calm when stuff broke at midnight. Coffee consumption skills also improved significantly.

What's next for Fly Dive

I want to expand this to other extreme sports—imagine wingsuit flying through canyons or scuba diving with whales. Adding VR support would be insane because then you'd actually feel surrounded by the experience. I'm also thinking about letting people customize everything: what they're wearing, where they're jumping, time of day, weather conditions. Maybe even create a community where people share their favorite dives. Long-term, it'd be cool to partner with actual adventure companies or use this for training simulations. Lots of possibilities. But, for now I will focus on JSON prompting because it is way more efficient than the traditional one.

Built With

• capcut