Kling

You describe a scene in text (or supply a start/end-frame image or reference video for Image-to-Video and Element Reference). A video diffusion model generates a 3-15 second clip by iteratively denoising it toward the prompt, synthesizing a matching native audio track - dialogue, effects, multiple languages - in the same pass rather than adding sound afterward. Multi-Shot mode either plans camera angles and scene transitions automatically or lets you specify each shot's framing and duration yourself, and Element Reference locks specific characters, objects or scenes so they stay consistent across shots.