Ai Test (personal project)
Duration: 5 hours
Image Generation: Chat GPT
Video Generation: Figma Weave
Seedance 2.0 Reference
Postworks: Adobe After Effects 2026
https://www.linkedin.com/feed/update/urn:li:activity:7473400948124651522/
Concept: A continuous AI-generated FPV flight from my apartment in the LES of Manhattan NYC to my grandma's house in South Korea, all guided by a single hand-drawn path.
For this project, I took a deep dive into AI workflows
and explored how these tools can expand storytelling and visual creativity.
Creative Process
1. Generate the wide map with ChatGPT.
via Figma Weave
Prompt: Remove all red lines, arrows, and guide marks from the final video. The markings are only used as camera motion guidance. Create a cinematic first-person FPV drone shot in one continuous take, following the path drawn on the image.
The journey begins inside a small apartment in Lower Manhattan, New York City. The drone exits through the window and accelerates through the city streets, weaving between skyscrapers t𝚑𝚎 𝚍𝚛𝚘𝚗𝚎 𝚛𝚊𝚙𝚒𝚍𝚕𝚢 𝚊𝚜𝚌𝚎𝚗𝚍𝚜 𝚊𝚛𝚘𝚞𝚗𝚍 𝚝𝚑𝚎 Washington Square Park. Then dives cleanly through the center of the Washington Square Arch, then flies toward the Statue of Liberty, smoothly orbiting the monument while revealing the New York skyline.
The drone accelerates across the Atlantic Ocean with cinematic speed ramps, flying low above clouds and ocean waves. As the journey continues, the environment gradually transitions into South Korea. The drone flies through the Seoul skyline at golden hour, passes over busy city streets, then 𝚏𝚕𝚒𝚎𝚜 𝚜𝚕𝚒𝚐𝚑𝚝𝚕𝚢 𝚊𝚋𝚘𝚟𝚎 𝚝𝚑𝚎 Seoul tower. Then it travels across peaceful countryside landscapes.
Finally, the drone arrives at Granmother's house in rural area by entrancing through the window.
Emotional storytelling, realistic FPV drone movement, cinematic speed ramps, natural camera inertia, seamless environment transitions, photorealistic lighting, grounded aerial realism, no cuts, no teleportation, smooth continuous motion, ultra realistic, no distortion, no duplicate buildings, no text, no watermark
Creative Challenges
Testing the new AI video model, Seedance 2.0
- Excellent prompt adherence
- Generates highly realistic environments and natural landscapes
- Produces smooth, cinematic camera movements
- High credit usage (expensive)
- Maximum clip length of 10 seconds, requiring multiple clips for longer scenes
Attempts