Whether you're filming a practical version or generating one with AI, a handful of avoidable mistakes account for most of the videos that don't quite land. Here's what to watch for.
Using a dark or blurry source photo
For AI-generated videos, a low-quality starting photo is the single biggest cause of a disappointing result. Take a moment to get a clear, well-lit, full-body photo rather than reusing whatever photo happens to be on hand.
Cropping out part of your body
Since the sequence involves your whole body — arms, torso, legs — a photo that cuts off your feet or crops in tight on your upper half gives the AI less to work with and can produce a less accurate result.
Overcomplicating the room
A cluttered, busy background pulls attention away from the actual illusion. Simpler is almost always better here.
Trying to physically replicate the stunt
As covered in our safety guide, there's no safe way to actually climb a ceiling, and you don't need to — the convincing versions are illusions, not real stunts.
Skipping a second look before posting
Before you post, watch your finished clip once all the way through. It's easy to miss a stray reflection, an odd shadow, or a background detail you'd rather not share until you watch the whole thing back.
Expecting instant results
AI generation can take a few minutes. Refreshing constantly or assuming something's broken after 30 seconds is a common frustration — give it time, and most tools will let you check back later rather than requiring you to wait on the page the whole time.
The fix for most of these
Almost every common mistake traces back to the source photo. Get that right — well-lit, full body, uncluttered background — and the rest tends to fall into place.
