Which Parts of a Startup Video Should Be AI-Generated?
Summary: emily jones discusses the creation of launch videos for a desk-booking app, emphasizing the importance of distinguishing between real and illustrative footage. They suggest labeling or rethinking shots that might mislead viewers about the app's features. Anthony Fox agrees, adding that AI is good for transitions and explanations but recommends using real footage for product demos to prevent confusion.
Imagine a desk-booking app’s launch video showing a sunny window seat. Someone opens the app to reserve it, but the workspace doesn’t exist. It was generated for the video.
I work on the site behind Seedance AI Video Generator. Here’s a simple idea for teams planning launch footage: add “What might viewers believe?” beside every shot.
For a desk-booking app, that helps separate two jobs:
Show what customers can use. Record the current booking flow with a demo account and fictional details. Film real locations that are available to reserve.
Explain the idea visually. Use generated supporting footage, such as an abstract animation of calendar blocks, where it won’t be mistaken for an actual listing or feature.
A workspace scene may suggest “I can book this desk.” A confirmation screen may suggest “My reservation is confirmed immediately.” Both expectations need to match the service.
That extra note in the shot list gives everyone reviewing the edit something concrete to check. If a scene needs an “illustrative footage” label, make it readable. If the label still leaves room for confusion, choose another shot.
Which parts of your launch video would you always capture yourself?
Muhammad Haris
The line that matters most here is asking what viewers might believe. A generated scene can suggest a feature that does not exist yet, and that small misunderstanding can follow a customer into their first real use. Keeping real product footage for anything a customer will rely on, and using generated material only where it cannot be mistaken for the product, is a clear boundary. At Onyx Growth Labs, we apply the same test to client messaging: if a visual could create an expectation the product cannot meet, it belongs in the idea pile, not in the launch video.
Anthony Fox
I think AI works well for transitions, background visuals, and explaining ideas. But for product demos, I would stick to real recordings so people can see how things actually work. It also helps to get a few people to watch the video before publishing and see if anything feels confusing or misleading.