Kling 4.0 brings a superpower: the ability to direct your scene frame by frame, with up to 10 keyframes in a single video. This opens up a lot of possibilities. Here are a few cases, and at the end a simulation of a real spot. How keyframes work You load a start frame, up to 8 middle frames and an end frame, in order. Each keyframe is a moment the video must reach: Kling generates the motion between one frame and the next, and lands on each image as you gave it. • The keyframes decide where the video goes: locations, framing, the state of the subject at each step. • The prompt decides how it gets there: the type of transition, the camera move, the actions between frames, the sound. • No prompt: @Kling_ai chooses the transitions on its own. Same keyframes, different prompt, completely different video. That is what the cases below show. The keyframes (Cases 1, 2 and 3) Five keyframes with the can standing in the center, same position and same size in every frame. Only the location changes: 1. Start: white studio 2. Middle: snow 3. Middle: sand 4. Middle: concert 5. End: rooftop at night Key: when you edit the middle keyframes, lock the can's position and size but let light, shadow and the contact with the ground adapt to each place. With everything forced to stay pixel-identical, the sand and snow looked fake.