klingai / kling-v2.6
kling-v2.6
Kling v2.6 is a unified multimodal model designed to solve the "silent video" problem of previous generations. Unlike v2.5, which focused on physical simulation, v2.6 focuses on sensory immersion. It understands the semantic link between visual motion and sound—automatically generating the sound of footsteps when a character runs, or lip-synced dialogue when a character speaks.
Kling v2.6 is a breakthrough multimodal generation model that introduces native audio-visual synthesis, capable of generating 1080p video with perfectly synchronized sound effects, dialogue, and background ambience in a single inference pass. Unlike previous iterations that required post-production for sound, v2.6 utilizes a unified Audio-Visual Diffusion Transformer architecture to understand the semantic relationship between motion and audio, delivering physically accurate Foley (e.g., footsteps, collisions) and precise lip-synced speech for a fully immersive, ready-to-publish cinematic output.
Example prompts
Great starting points for kling-v2.6.
API quick start
Run kling-v2.6 with a single API call.
{
"prompt": "The man in the orange hat and blue jacket…",
"negativePrompt": "...",
"duration": 5,
"sound": "off"
}