AI Water Slide Video Generator - POV Down an Impossible Drop
Impossible Slide makes the first-person waterslide clip that broke out on Sora in 2025: you pick where the slide drops from, and the video is one unbroken ten-second action-camera take from the rider's own eyes - sitting at the mouth of a clear tube with the ground visible through its floor, pushing off, riding a banked serpentine that rolls the horizon past vertical, and getting thrown clear into the water at the end. The camera never cuts to an outside view of the slide, which is the difference between a ride and a flyover of a water park.
What is a POV water slide video?
It is a single continuous clip filmed from the rider's own head on a wide-angle action camera, going down an enclosed transparent tube that is far too high to exist - you can see the ground through the floor the whole way, and the tube banks and rolls the horizon on its way to the water.
The real ones are action-camera footage from water parks. The viral ones are generated, and they work because they keep the grammar of the real ones: the fisheye curve at the edges, the rider's own legs in the bottom of the frame, the shake, and the moment the camera goes under the water and comes back up. What makes them frightening rather than merely fast is that the tube is clear, so the drop never leaves the shot.
What changes between clips is only where the slide is and what is at the bottom of it, which is why this page asks one question and commits everything else.
Where can the slide drop from?
Eight settings: the side of a skyscraper at dusk, a water park's tallest tower, a volcano flank, an alpine cliff, a cruise ship, a desert canyon, a jungle waterfall or the face of a dam. You can also type your own.
Each option brings its own light, its own noise and its own water at the bottom - a lit city grid under the floor at dusk is a different clip from white sun over a water park at midday, and both are in the list rather than behind a second question. The tube, the bends and the camera are the same every time.
A typed drop works the same way, with one rule the prompt is built around: it has to end in water, because the last second of the video is the splash and the camera coming back up through the surface.
Does it have sound?
Yes, and it is generated in the same pass as the picture. It echoes the way a closed tube echoes, rises through every bend, peaks as the rider is thrown clear at the end, and goes dull for the second spent under the water before surfacing.
There is no music and no captions - running water, wind, four shouted lines from the rider and the splash. Sound cannot be turned off on this page: silent, the landing is white foam and nothing.
How long is the video and what does it cost?
Ten seconds by default, vertical 9:16, for 12 tokens at 720p or 6 at 480p. Twenty seconds is in Advanced and costs 24 tokens at 720p.
Both lengths have their own beat sheet rather than one being a trimmed copy of the other. The ten-second cut keeps the two seconds at the mouth and the two in the water whole and spends its middle on one left-right pair of bends; the twenty-second cut buys a second pair, a straight falling section between them and a fourth shouted line. The bends are the format, so the short version gives up everything else before it gives up banking.