Kling 4.0 结束预览，全量开放：30 秒单段、10 关键帧、15 参考素材
Kling 4.0 于 10 月 16 日向所有用户开放，单段最长 30 秒、10 张关键帧、15 项多模态参考，支持九种语言口型与双声道，提示词约 8000 token，多镜头可延至约 2 分钟，MCP/CLI 已接入，4K 与 10-bit HDR 待上线。
People kept asking when Kling 4.0 would leave preview. As of today, it is live for all users on Kling's site. Kling AI opened the full model on October 16 (GlobeNewswire and Pandaily both covered the rollout; Kling's own account posted that it was fully available). You can generate a single native clip up to 30 seconds, steer it with up to 10 keyframes, and feed as many as 15 multimodal references, including images, video, and audio. Dual-channel stereo and lip sync cover Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French, and Hindi, plus several Chinese dialects and English accent variants. Prompt input goes up to about 8,000 tokens, and multi-shot continuation can extend a sequence toward about two minutes. MCP and CLI hooks are wired in for agent and developer workflows. 4K and 1080p 10-bit HDR output, plus an enterprise API, are listed as coming soon. Adobe says Kling 4.0 is also coming to Firefly. For a classroom or studio drill this week: pick one 20-second dialogue scene, lock three keyframes (start, beat change, end), and generate picture and audio together. Watch once with sound, then mute the clip and check whether the mouth still lands on the words. 🎬 #AIFilmmaking Link in Comments