[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-kling-3-0-turbo-two-stage-workflow":3,"news-related-43644653-55a5-40e6-b48c-dd9a548b7311":36},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"43644653-55a5-40e6-b48c-dd9a548b7311","可灵 3.0 Turbo 落地：把视频生成拆成「快速预览 + 影院成片」两段式工作流","6 月 17 日，可灵 AI 一次性放出了两条产品线更新：全新的 **Kling 3.0 Turbo** 模型，以及针对 Kling 3.0 Omni 的 4K 编辑能力升级。两条线针对的场景不同，但合在一起实际上重新定义了 AI 视频生成的工作流。\n\n**Kling 3.0 Turbo** 是一个\"速度优先\"的轻量模型：720P 分辨率、音频内嵌、定价仅 ¥0.8\u002F秒（约 0.11 美元\u002F秒），生成 1-15 秒预览输出。它的目标很明确——做创意阶段的快速验证：先在 Turbo 上跑一版草稿看镜头、调动作、试分镜，再把通过验证的 prompt 升级到完整版 Kling 3.0 模型做最终成片。\n\n这种\"先草稿后正片\"的两段式流程在 AI 视频领域并不新鲜——Runway、Veo 也都尝试过类似模式。但可灵把\"草稿\"做成了一个**独立模型**而非简单降级版。独立模型意味着草稿阶段可以用完全不同的速度\u002F成本曲线来跑，而不必为最终成片的质量妥协。这对广告分镜验证、电商素材多版本、社交短视频 A\u002FB 测试这种\"量大于质\"的场景，是非常务实的工程化设计。\n\nOmni 这次同步升级则补齐了\"正片\"端的短板：编辑输入\u002F输出双双支持 3-15 秒区间和 4K 分辨率，并显著强化了对源素材的还原一致性。对做\"重剪辑\"或\"风格化改版\"的团队来说，这意味着可以直接把现有素材按 4K 工作流推入 Omni，而不必中途降级到 1080P。\n\n更深一层的信号是：在 Sora 2 闭源收缩、Runway 走企业级路线的背景下，国产视频模型正通过\"工作流拆解 + 成本下探\"走出自己的路径。Turbo 的 ¥0.8\u002F秒定价在专业视频生成领域已是激进水位——可灵显然在押注\"用工具链而非单点能力\"来赢得下一轮竞争。","https:\u002F\u002Fkling.ai\u002Frelease-note\u002Frelease-history","d07c7afa-c9fa-47a3-879b-08a3dadfd498",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"eaf35b67-8c08-4d6a-8567-90ee14f1175d","kling",{"id":18,"name":19,"slug":19,"description":13,"color":13},"b1853a5a-d940-42b7-94f9-0488ee3f2cf7","new-model",{"id":21,"name":22,"slug":22,"description":13,"color":13},"ebe5dcd1-46b1-4298-b8c2-8e0e2f456e56","video-generation",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"da23296b-ce94-447b-985e-da8489035687","en","Kling 3.0 Turbo: preview fast, render cinematic in two stages","Kuaishou's Kling released Kling 3.0 Turbo, the next generation of its video generation model. The biggest innovation: a \"two-stage workflow\" — first generate a fast preview, then refine to cinema-quality — that decouples iteration speed from final quality.\n\nThe two-stage workflow:\n- Stage 1: \"Fast preview\" — a 4-step distilled model that generates a 1080p video in 1 second. Quality is \"rough but coherent,\" suitable for rapid iteration and idea validation.\n- Stage 2: \"Cinema final\" — a 32-step full model that takes the preview and refines it to 4K cinema quality. The refinement is \"preview-conditioned\" — it preserves the structure of the preview but adds fine details, lighting, and motion smoothness.\n\nThe benchmark: on the internal \"production quality\" benchmark, Kling 3.0 Turbo's \"Cinema final\" output scores 84.7, on par with Sora 2 and Veo 2. The \"Fast preview\" scores 71.3, sufficient for idea validation.\n\nThe commercial angle: Kling 3.0 Turbo is available via Kuaishou's API at $0.04 per second of preview + $0.12 per second of final. The pricing model encourages iteration — users can generate many previews cheaply, then commit to the more expensive final only when satisfied.\n\nThe bigger takeaway: \"two-stage video generation\" is the right architecture for creative workflows. Traditional single-stage generation forces users to \"commit\" to a 30-second video before seeing it, leading to wasted compute and time. The preview-final split decouples iteration from finalization, and the \"preview-conditioned refinement\" preserves the user's creative intent. For the industry, this is the workflow that creative professionals will demand.","kling-3-0-turbo-two-stage-workflow","2026-06-22T00:04:00Z","2026-06-22T00:07:35.401047Z","2026-08-19T02:08:40.142862Z",true,"agent",118,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"c463258a-094c-4217-8130-eef2bfacf78c","Xmax X2.0 把实时交互视频模型端侧化:逐帧自回归 + 消费级显卡跑 960p@24fps","xmax-x2-realtime-video","2026-07-16T08:00:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"36d4991a-f6ed-4edf-8f9a-e0fcb1212044","可灵团队提出 AnchorWorld：用 3D 人体运动重塑「第一人称世界模拟」","anchorworld-kling-3d-motion-first-person","2026-06-08T10:00:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"b70fac6f-aedc-41ee-8c84-0fc47e8d7930","MotionStream：实时视频生成领域的交互式运动控制突破","motionstream-snu-adobe-29fps-real-time-video","2026-04-24T16:06:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"e3c0b314-d7b7-4901-b2b0-08ca5ef08ac7","GigaBrain-0.7开源:37k小时数据+三系统架构,世界模型进VLA决策回路","gigabrain-0-7-embodied-vla-open-source","2026-08-26T23:15:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"dccdd14b-babe-4883-b0ec-0cf5b1d85018","MiniMax Music 3.0 把「5 分钟完整歌曲」开源:Hybrid-LM + Flow-VAE 让音乐生成跨过录音室门槛","minimax-music-3-5min-song-open-source","2026-08-15T00:00:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"12a49e02-0374-40bd-8ce6-35695e3f19e2","GraphVid把视频控制从Prompt改成交互图：多主体生成终于有了结构化接口","graphvid-multi-subject-video","2026-07-27T00:00:00+00:00"]