Kling image to video: how to get the best clip

Your image already shows what things look like. The prompt should say what happens: what the subject does, and what the camera does while it does it. Only then the surroundings.

Kling runs to fifteen seconds, so give it a shape: say what happens at the start and what has happened by the end. Otherwise you get fifteen seconds of one gesture.

The aspect ratio comes from your start image rather than from a setting, so the frame is decided before you generate. Crop the still to 9:16 for a Reel or 16:9 for a widescreen shot first.

Kling 3.0 makes audio in the same pass, with more than one speaker. Name a character with an element tag and it keeps that character through the clip. And it takes several prompts in one generation for a clip with more than one scene.

Audio is native in Chinese and English. Other languages are translated to English, so a line in Dutch is not something to plan around.

The Kling family has many tiers, and they are not only faster or slower: the Pro tiers are visibly better. Draft on a cheaper tier, then render the take you keep on Pro. Every tier shows its price on the button.