Image-to-video prompts that move
Kling's own formula for animating a still: name the subject, describe its motion, and keep it simple.
The image already does half the work
When you animate a still in Kling, the picture supplies the scene, the look and the lighting. Your prompt only has to say what moves. Kling's quickstart boils it down to subject plus movement, with an optional background plus movement. So instead of describing the photo again, write something like: the woman puts on her sunglasses, and a ray of light appears behind her.
Be concrete about the action
Vague instructions are often ignored. If you only write a two-word action, the model may fall back to its own default motion, which on illustrated or painted images is often a slow pan or barely any movement. Say who does what, with which hand, in which direction. For several subjects, give each one its own action in order.
Keep motion believable
Use simple words and short sentences. Ask for motion the image could plausibly show. If your description drifts too far from what is in the picture, Kling may insert a cut or a transition to get there. Complex physics, such as a bouncing ball or something thrown high into the air, is still hard to get right, so avoid building a clip around it.
Text-to-video uses the same logic
Without a start image you add a short scene description: subject with a few visual details, one action that fits a few seconds, the setting, then optional camera, lighting and mood. Kling's guide also warns that exact counts of people or objects are unreliable.
Set it up for short-form
Choose 9:16 for TikTok, Reels and Shorts. Start from a clean, well-lit image with one clear focal point, write one action, and generate a few variations. Small changes in wording often give very different motion, so keep the prompt that worked and change one thing at a time.