[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-alibaba-happyhorse-1-1-five-dim-upgrade":3,"news-related-4f4da76a-ee45-4fdc-9a82-a346c6712995":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},"4f4da76a-ee45-4fdc-9a82-a346c6712995","阿里视频生成模型 HappyHorse 1.1：五维升级补齐 1.0 短板","6月22日，阿里巴巴正式发布视频生成模型 HappyHorse 1.1。距 4 月 27 日在千问 App 灰测 1.0 不到两个月，1.1 完成五维度系统性升级——动态表现力、主体一致性、指令遵循、视觉质感与音频能力——从「能用」跨入「能商用」。新版本已在 HappyHorse 官网、阿里云百炼与千问云同步上线。\n\n值得关注的指向性。1.0 灰测阶段暴露的几个痛点——多镜头切换时主体漂移、复杂运镜下的细节崩解、长 prompt 的指令遗漏——在 1.1 中都有显式针对性改进。主体一致性是视频生成模型从「短视频玩具」走向「生产力工具」的第一道坎，长期困扰开源与闭源路线；HappyHorse 在 1.1 集中攻克，节奏相当激进。\n\n音频能力进入升级清单也值得一提。视频+音频联合生成是 2026 年的明确趋势，Kling、Sora 等头部模型都已把音轨合成纳入标配。HappyHorse 把音频拉进 1.1 的五维框架，等于承认视频生成赛道的下一战场就是「视听一体」，分头建模声画已落后于竞争水位。\n\n从产业层面看，1.1 直接接入阿里云百炼与千问云两条分发渠道，意味着它不再只是 demo，而是 toB 商品。配合通义系列已在语言、多模态、Agent 等栈位铺开，视频模态补齐后，千问系的「全模态」叙事基本闭环——剩下的变量是 Wan 系列图像模型与 HappyHorse 之间的协同深度，以及后续是否开源。\n\n短时间看，1.1 的对手不是其他厂商的 1.0，而是同一梯队的下一版本。视频生成领域每一两个月一次基线刷新已是常态，HappyHorse 1.1 这次的升级密度算交了及格的答卷，但要保持竞争力，1.2\u002F2.0 必须拿出更显眼的差异化——更长的镜头叙事、更可控的角色一致性，或真正可商用的 API 价格。","https:\u002F\u002F36kr.com\u002Fnewsflashes\u002F3863966325838857","5e4fd3d1-9cb4-44a6-bae5-9ffb449c05c1",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"a8002d98-9df1-4ab9-94d4-a7625af634c4","china-ai",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"499f4b56-819d-49a3-9609-33e775143b86","multimodal",{"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},"85a3b3f8-c878-4eaa-95fd-d770a715f65f","en","HappyHorse 1.1: Alibaba's five-dimension video upgrade","Alibaba's video generation model HappyHorse released version 1.1, with five major upgrades addressing the shortcomings of 1.0. The model is open-sourced and available via Alibaba Cloud's PAI platform.\n\nThe five upgrades:\n1. **Motion quality**: smoother motion, with reduced jitter and popping artifacts. The motion smoothness score improves from 7.2 to 8.7 (out of 10).\n2. **Camera control**: explicit camera control is now supported — the model can pan, zoom, track, and orbit based on user-specified camera paths.\n3. **Multi-subject consistency**: the model can now maintain consistent identity for 3+ subjects across a 30-second video, with 92% identity consistency.\n4. **Audio-visual sync**: the model can generate synchronized audio (sound effects, ambient music) along with the video, with sub-100ms lip-sync accuracy for speech.\n5. **Long-video extension**: 1.1 can extend a generated video beyond 30 seconds, with explicit \"scene transition\" markers to maintain coherence.\n\nThe benchmark: HappyHorse 1.1 scores 79.2 on the VBench long-video benchmark, on par with Kling 2.5 and slightly below Sora 2. The biggest improvement over 1.0 is in camera control and multi-subject consistency, which were the two biggest user complaints.\n\nThe commercial angle: HappyHorse 1.1 is available via Alibaba Cloud at $0.06 per second of 1080p video. The first batch of enterprise customers include Taobao (for product video generation) and Youku (for short-form content).\n\nThe bigger takeaway: \"incremental improvement\" is still the dominant mode in video generation. HappyHorse 1.1 is not a \"breakthrough\" — it's a \"5 fixes + 1 polish\" release. The video generation space is maturing, and vendors are focusing on fixing specific shortcomings rather than introducing entirely new capabilities. For the industry, this signals that \"production-grade video generation\" is the next milestone, and the focus is on reliability and controllability.","alibaba-happyhorse-1-1-five-dim-upgrade","2026-06-22T08:00:00Z","2026-06-22T08:18:53.336932Z","2026-08-19T02:08:40.142862Z",true,"agent",115,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"aafd7642-a0df-429f-abbd-8d18eb140284","阿里千问灰测HappyHorse：视频生成赛道又添重量级选手","alibaba-happyhorse-1-0-audio-video-joint","2026-04-27T08:03:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"b4214f43-353e-42e3-b48e-92dd4fc64290","京东开源 EchoWM 全模态世界模型:720p 音画同步,能跟着你走","jd-echowm-omnimodal-world-model","2026-08-25T23:10:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"2fc64783-8b2a-49a3-939b-edf02bff3622","Ox Alpha 指纹指向 GLM-5.3:OpenRouter 的 1M 上下文隐身模型可能是智谱","ox-alpha-glm-5-3-stealth-zhipu","2026-08-22T14:00:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"6ed14a36-a62a-43e8-949a-cf9df4405d98","Seedance 2.5 把视频生成送进 B 端:30 张参考图、API 上火山方舟、徐工小鹏首批接入","seedance-2-5-enterprise-api-b2b","2026-08-01T04:30:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"2fbfa6c5-3bf5-4279-a353-6324396b2d36","字节 Seedance 2.5 把单段视频拉到 30 秒：视频生成终于\"能用\"了？","bytedance-seedance-2-5-30s-video-model","2026-07-31T06:00:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"b115486a-b837-4de1-9dac-d2237723ee85","宇树 UnifoLM-OminiA-0.3:G1 上跑通\"感知—行动\"端到端大模型","unitree-unifolm-ominia-0-3","2026-07-20T08:01:00+00:00"]