Project Story: AI Video Overlay for TikTok

Inspiration

We were inspired by the rapid evolution of AI-native interfaces. Today, people no longer just “consume” content; they co-create it with AI. TikTok already empowers millions of creators, and we wanted to imagine:
What if anyone could generate and edit videos with just words?

This idea came from watching how Gen Z and younger audiences remix and iterate on trends at lightning speed. Traditional video editing feels slow and technical — so we set out to build a prompt-driven video overlay that lets creativity flow as naturally as chatting with an AI.

How We Built It

We built the project with Lynx, TikTok’s open-source cross-platform UI technology, allowing us to deliver one codebase across iOS, Android, and Web.

• UI/UX: Built in React-style declarative JS with Lynx, including a timestamp input system and video preview panel.
  
• AI Video Generation: Connected to a generative video API (diffusion/synthesis-based), returning two candidate outputs for user choice.
  

What We Learned

• How Lynx accelerates UI iteration across platforms without duplicating work.
  
• The importance of human-in-the-loop AI design — giving users control over generated outputs (instead of a single result).
  
• Prompt engineering for video is less about single instructions and more about chaining context at different moments in time.
  
• Balancing creative freedom with usability: too many options overwhelm users, too few limit them.
  

Challenges We Faced

• Iterative edits: Regenerating only part of a video while keeping the rest consistent required experimentation with conditioning strategies.
  

Takeaway

In the end, we created an AI-powered TikTok overlay that makes video editing as simple as chatting with an AI.
It represents our vision of how UI/UX will evolve in the AI era — fast, creative, iterative, and human-centered.

Built With

• lynx