可灵（Kling）发布 4.0 视频生成模型，新增立体声与多语言唇形同步，另推 Flash 轻量版
Kling AI drops version 4.0 with stereo audio, multilingual lip sync, and a lighter Flash variant
- 1可灵发布 4.0 版视频生成模型，单次最长可生成 30 秒，提示词理解扩展到 8000 tokens
- 2新增双声道立体声，唇形同步覆盖十余种语言，Omni Reference 单次最多接受 15 个参考素材
- 3第三方对比称 Kling 4.0 与 Wan 3.0、Seedance 2.5 相当但未领先，进步速度或正放缓
- 4Kling 4.0 Flash 限 720p、20 秒、更快更省，支持经 MCP 和 CLI 接入智能体工作流
可灵 4.0 把重点放在素材可控性与工作流接入上，第三方对比显示头部视频模型差距收窄，能力提升节奏开始放缓。
- 影响：多语言唇形同步和立体声让带配音的营销视频可直接生成，广告与本地化素材可先小规模试产；参考素材机制利于保持角色与产品一致性；API 定价未公布，接入成本要等公布后再核算。
- 参考：可考虑用同一组提示词与素材在 Kling 4.0、Wan 3.0、Seedance 2.5 上各跑一轮对比再选型；值得先测算 Flash 版在草稿与内部素材场景的省费空间；等 API 定价公布后再评估是否用 MCP 搭自动化流水线。
- 依据：功能与参数均为可灵官方口径，暂无独立验证；与 Wan 3.0、Seedance 2.5 的对比出自 The Decoder 的自评视频，结论是打平但未领先。
AI video generator Kling AI has released version 4.0 with higher resolution, more precise controls, and a fast Flash variant that the company says should close the gap between AI-generated and professional video.
Users can now generate up to 30 seconds of video in a single pass, and the system accepts up to ten keyframe images as input to pin down scene transitions, character states, and story beats. Prompt comprehension has been expanded to 8,000 tokens.
The upgraded "Omni Reference" system accepts up to 15 reference assets in a single generation, according to Kling AI, including up to 10 images, 5 videos with a combined maximum of 30 seconds, and 7 elements. Storyboards, wireframes, image grids, white-model videos, and depth-map videos all work as references.
A "Creative Recreation" feature lets users take existing videos as templates and swap out products or characters. The company says individual elements like facial expressions, body movements, camera angles, or backgrounds can be tweaked without affecting the rest of the clip.
The video below compares Kling 4.0 against Wan 3.0 and Seedance 2.5, both from Chinese competitors. Kling 4.0 holds its own but doesn't pull ahead, especially when it comes to logical shot sequencing, physics, and natural-sounding dialogue. That could be an early sign that the breakneck pace of progress in AI video models is starting to level off, though the next Wan and Seedance releases will tell us more.
Stereo audio and multilingual speech round out the update
Kling 4.0 adds two-channel stereo audio and improved lip sync across more than ten languages, including German, English, Japanese, Korean, and Spanish, along with various dialects and accents, according to the company. In-video text generation now works in multiple languages too and stays stable during camera movements. 10-bit HDR output at 4K and 1080p and a video extension function that would stretch clips to up to two minutes are both listed as "coming soon" by Kling AI.
Alongside the main model, Kling AI is shipping a lighter Flash variant. Kling 4.0 Flash is designed for faster generation at lower cost, though it tops out at 720p and 20 seconds of video. The creation interface has been redesigned, and Kling 4.0 can now be accessed through MCP and CLI, which makes it possible to build automated workflows with AI agents. Pricing for the Kling 4.0 API hasn't been announced yet, with support listed as "coming soon."
AI News Without the Hype – Curated by Humans
Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.
来源：The Decoder：AI News · the-decoder.com