Kling 4.0 launches with 30-second clips, stereo audio and a cheaper Flash variant
Kling AI released Kling 4.0 on the evening of October 15, according to PingWest and The Decoder. The update generates up to 30 seconds of video in a single pass, accepts up to 15 reference assets (including up to 10 keyframe images) through Omni Reference, raises prompt input to 8,000 tokens, and adds two-channel stereo audio, improved lip sync across more than ten languages and in-video multilingual text; 21:9 framing and 10-bit HDR at 4K and 1080p are listed as coming soon. A lighter Kling 4.0 Flash targets fast, low-cost iteration but tops out at 720p and 20 seconds, and the model is also reachable through MCP and CLI, with API pricing not yet announced. PingWest's hands-on found stronger fast motion, close combat and emotional performance, while The Decoder said a comparison video shows Kling 4.0 roughly level with Wan 3.0 and Seedance 2.5 rather than clearly ahead.
Kling AI is separately funded at a $18 billion post-money valuation after a roughly $3 billion raise from Kuaishou, and Kling 4.0 is the version it is pitching at professional film, advertising and post-production workflows, per TMTPost.
盼望着、盼望着，可灵终于带着它的最新模型来了。
10 月 15 日晚，可灵 AI 的 Kling 4.0 正式全量上线，给已经激烈焦灼的视频模型战场又加了一把火。
和今年下半年大部分视频模型一样，Kling 4.0 也瞄准了专业用户：强化大幅动态和复杂运镜，将全能参考扩充至最多 15 项素材、关键帧扩充至最多 10 张图片，单次原生生成也延长到了 30 秒。
动作更复杂、控制更精准、叙事更完整，是这场竞争共同的方向。
但看完整份更新清单，会发现可灵这次的劲头格外足：专业用户想要的，它几乎都想安排上。
不仅输入端，文本长度上限提高到 8000 tokens，可以一次容纳更细致的分镜脚本；输出端，21:9 宽画幅和 4K 和 1080p 的 10-bit HDR，进一步照顾构图与后期需求。还有面向高频创作的 Flash 版本，为反复试稿提供了更快、更具性价比的选择。此外，Kling 4.0 API 也将于近期上线。