[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-midjourney-public-beta-visual-generation-history-en":3,"article-related-midjourney-public-beta-visual-generation-history-en":25,"series-industry-7e97034b-97b7-4bd6-86dd-c0267fefd4ed":74},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":11,"views":22,"created_at":23,"published_at":24,"topic_cluster_id":11},"7e97034b-97b7-4bd6-86dd-c0267fefd4ed","midjourney-public-beta-visual-generation-history-en","Midjourney公测背后的视觉生成史","\u003Cp>7月，\u003Ca href=\"https:\u002F\u002Fwww.midjourney.com\" target=\"_blank\" rel=\"noopener\">Midjourney\u003C\u002Fa>进入公测，创始人 \u003Ca href=\"https:\u002F\u002Fwww.linkedin.com\u002Fin\u002Fdavidholz\u002F\" target=\"_blank\" rel=\"noopener\">David Holz\u003C\u002Fa> 没有把产品做成传统 App，而是把入口放进了 \u003Ca href=\"https:\u002F\u002Fdiscord.com\" target=\"_blank\" rel=\"noopener\">Discord\u003C\u002Fa>。这个选择很聪明：用户不是一个人对着空白画布，而是在一个公开频道里看着别人不断生成、修改、再生成，像在围观一场实时创作秀。\u003C\u002Fp>\u003Cp>这种“广场式”体验迅速放大了传播效率，也让 Midjourney 的审美标签变得非常鲜明。它的图像不追求机械式还原，更像是把“好看”写进了默认参数里，尤其是 V-series 之后，那种偏 CG、偏海报、偏概念设计的质感，几乎成了它的招牌。\u003C\u002Fp>\u003Cp>如果把这件事放回技术史里看，Midjourney 只是最新一轮爆发。视觉生成已经走了七十多年，从早期的规则绘图，到神经网络，再到今天的大模型扩散生成，今天我们看到的“点几下就出图”，其实是几代研究和产品路线叠加后的结果。\u003C\u002Fp>\u003Ch2>Midjourney为什么先赢在Discord\u003C\u002Fh2>\u003Cp>Midjourney 早期没有把精力放在独立客户端上，而是直接押注 Discord。这个决定降低了使用门槛，也把生成过程变成了社交内容本身。用户发一句提示词，几秒后就能得到四张图，再继续放大、重绘、变体，整个过程天然适合围观和转发。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775171754306-tnkj.png\" alt=\"Midjourney公测背后的视觉生成史\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>对生成式产品来说，分发方式往往和模型能力一样重要。Midjourney 的做法把“使用”变成了“展示”，把“结果”变成了“话题”。这也是它比很多同类工具更快出圈的原因之一。\u003C\u002Fp>\u003Cp>它的审美策略也很明确。Midjourney 不太执着于照片级真实感，而是持续强化一种更容易被普通用户接受的视觉风格：高对比、强光影、细节饱满、构图完整。对设计师来说，这意味着它更像一个灵感机器；对普通用户来说，它更像一个“自动出片”的工具。\u003C\u002Fp>\u003Cul>\u003Cli>入口在 Discord，降低了安装和学习成本\u003C\u002Fli>\u003Cli>默认生成结果更偏艺术化，而非纯写实\u003C\u002Fli>\u003Cli>公开频道让每次生成都带有社交传播属性\u003C\u002Fli>\u003Cli>V-series 强化了统一审美，形成明显品牌辨识度\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>从规则绘图到扩散模型\u003C\u002Fh2>\u003Cp>视觉生成不是最近几年才出现的想法。早在 20 世纪中期，研究者就已经在尝试用程序生成图形，只是那时的方法更接近“手工写规则”。计算机能画线、画几何图案、做简单变形，但离今天这种“理解提示词并生成完整图像”还很远。\u003C\u002Fp>\u003Cp>真正把这条路线推向实用的是深度学习。2014 年，Ian Goodfellow 提出了 \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F1406.2661\" target=\"_blank\" rel=\"noopener\">GAN\u003C\u002Fa>，生成图像第一次有了更强的逼真感。随后，扩散模型开始接管高质量生成任务，\u003Ca href=\"https:\u002F\u002Fopenai.com\u002Findex\u002Fdall-e-2\u002F\" target=\"_blank\" rel=\"noopener\">OpenAI\u003C\u002Fa> 的 \u003Ca href=\"https:\u002F\u002Fopenai.com\u002Findex\u002Fdall-e-2\u002F\" target=\"_blank\" rel=\"noopener\">DALL·E 2\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fstability.ai\" target=\"_blank\" rel=\"noopener\">Stability AI\u003C\u002Fa> 的 \u003Ca href=\"https:\u002F\u002Fstability.ai\u002Fstable-diffusion\" target=\"_blank\" rel=\"noopener\">Stable Diffusion\u003C\u002Fa>，把“文字到图像”的能力真正带到了大众手里。\u003C\u002Fp>\u003Cp>Midjourney 的差异不在于它发明了生成图像这件事，而在于它把模型输出包装成了一种稳定的审美体验。很多模型能生成“正确”的图，但 Midjourney 更擅长生成“愿意发出去”的图。\u003C\u002Fp>\u003Cblockquote>“The future of AI is not about replacing humans, it’s about amplifying human creativity.” — David Holz\u003C\u002Fblockquote>\u003Cp>这句话常被拿来解释 Midjourney 的产品哲学。它并没有把自己定义成一个替代设计师的工具，而是把重点放在创意放大上。这个方向也解释了为什么它会优先优化风格、构图和整体观感，而不是一味追求像素级还原。\u003C\u002Fp>\u003Ch2>四个关键节点看视觉生成的演进\u003C\u002Fh2>\u003Cp>如果把视觉生成史压缩成几个节点，会更容易看清 Midjourney 为什么会在这个时间点爆发。每一代技术都在解决前一代的短板，而用户能感知到的，往往是结果而不是算法细节。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775171753485-f7ht.png\" alt=\"Midjourney公测背后的视觉生成史\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>先看几个具体数字。\u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F1406.2661\" target=\"_blank\" rel=\"noopener\">GAN 论文\u003C\u002Fa>发表于 2014 年；\u003Ca href=\"https:\u002F\u002Fopenai.com\u002Findex\u002Fdall-e-2\u002F\" target=\"_blank\" rel=\"noopener\">DALL·E 2\u003C\u002Fa> 在 2022 年把文字生成图像带到更高分辨率；\u003Ca href=\"https:\u002F\u002Fstability.ai\u002Fstable-diffusion\" target=\"_blank\" rel=\"noopener\">Stable Diffusion\u003C\u002Fa> 同年开源后迅速扩散到本地部署和第三方应用；\u003Ca href=\"https:\u002F\u002Fwww.midjourney.com\" target=\"_blank\" rel=\"noopener\">Midjourney\u003C\u002Fa> 通过 Discord 先做社区，再做产品。\u003C\u002Fp>\u003Cul>\u003Cli>2014：GAN 让生成图像第一次具备较强真实感\u003C\u002Fli>\u003Cli>2022：DALL·E 2 把文本到图像的质量推到新高度\u003C\u002Fli>\u003Cli>2022：Stable Diffusion 开源后迅速进入开发者和创作者工作流\u003C\u002Fli>\u003Cli>Midjourney：用 Discord 社区把生成过程变成传播内容\u003C\u002Fli>\u003C\u002Ful>\u003Cp>这条链条说明一件事：视觉生成的竞争早已不只是“谁的模型更强”，而是“谁能把模型变成用户每天都会打开的产品”。Midjourney 在这一点上做得很早，也做得很准。\u003C\u002Fp>\u003Ch2>OpenAI为何关停Sora的讨论\u003C\u002Fh2>\u003Cp>标题里提到“OpenAI 为何关停 Sora”，但更准确地说，\u003Ca href=\"https:\u002F\u002Fopenai.com\u002Fsora\" target=\"_blank\" rel=\"noopener\">Sora\u003C\u002Fa> 讨论的是视频生成的边界，而不是单纯的产品成败。OpenAI 公开展示 Sora 时，重点放在长时序一致性、复杂场景和镜头运动上。它让外界第一次清楚看到，视频生成已经从“短片段演示”走向“可叙事的镜头语言”。\u003C\u002Fp>\u003Cp>但视频比图片难得多。图片只需要在一个瞬间成立，视频则要在时间轴上保持人物、物体、光线和运动逻辑一致。生成一张漂亮的图像已经不容易，生成十几秒还不崩的画面，更像是在和物理规律、记忆一致性、镜头调度同时较劲。\u003C\u002Fp>\u003Cp>这也是 Midjourney 和 Sora 的分野。Midjourney 把注意力放在静态图像的审美稳定性上，Sora 则把问题推进到动态世界建模。一个解决“好看”，另一个解决“会动且说得通”。\u003C\u002Fp>\u003Cp>从产品角度看，这两条路线都说明生成式 AI 已经过了单纯拼参数的阶段。接下来比的，是谁能把能力做成稳定的工作流，谁能让创作者愿意把日常任务交给它。\u003C\u002Fp>\u003Ch2>接下来谁会更吃香\u003C\u002Fh2>\u003Cp>接下来真正有竞争力的产品，未必是“最像”的那个，而是“最适合某种创作场景”的那个。Midjourney 已经证明，审美一致性和社区传播能让一个模型迅速破圈；Sora 则提醒大家，视频生成的门槛高得多，谁先解决长时序一致性，谁就更接近生产级应用。\u003C\u002Fp>\u003Cp>对开发者和产品经理来说，这里有个很现实的判断标准：模型能力只是起点，入口设计、反馈速度、审美策略、版权边界、工作流整合，都会直接影响最终结果。单纯把 API 暴露出来，已经不够了。\u003C\u002Fp>\u003Cp>如果你想判断下一波视觉生成产品谁会跑出来，可以盯住这些指标：\u003C\u002Fp>\u003Cul>\u003Cli>生成结果的稳定性，而不是单次演示的惊艳程度\u003C\u002Fli>\u003Cli>社区传播效率，尤其是是否天然适合分享\u003C\u002Fli>\u003Cli>是否能嵌进设计、广告、短视频和电商的日常流程\u003C\u002Fli>\u003Cli>对风格控制和版权风险的处理方式\u003C\u002Fli>\u003C\u002Ful>\u003Cp>Midjourney 的故事说明，生成式 AI 的胜负手经常不在模型参数表里，而在用户第一眼看到的那张图里。下一阶段，谁能把“好看、可控、可复用”同时做好，谁就更可能拿到真正的生产力入口。问题已经不是图能不能生成，而是谁会把生成结果变成自己的工作标准。\u003C\u002Fp>","Midjourney在Discord公测后走红。它的审美偏好算法和社交式交互，改写了图像生成的传播方式。","zhuanlan.zhihu.com","https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F2020798020871042635",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775171754306-tnkj.png","industry","en","93bd4b46-8be3-4118-bb94-6e230bf4bc7d",[17,18,19,20,21],"Midjourney","Discord","视觉生成","DALL·E 2","Stable Diffusion",8,"2026-04-02T23:15:37.853381+00:00","2026-04-02T23:15:37.737+00:00",{"tags":26,"relatedLang":33,"relatedPosts":37},[27,29,31],{"name":17,"slug":28},"midjourney",{"name":18,"slug":30},"discord",{"name":21,"slug":32},"stable-diffusion",{"id":15,"slug":34,"title":35,"language":36},"midjourney-public-beta-visual-generation-history-zh","Midjourney公測背後的視覺生成史","zh",[38,44,50,56,62,68],{"id":39,"slug":40,"title":41,"cover_image":42,"image_url":42,"created_at":43,"category":13},"96ad3567-ab75-487a-b9ac-656da06056ef","deepmind-veterans-are-leaving-london-en","DeepMind老兵正在离开伦敦","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782777770669-33e7.png","2026-06-30T00:02:29.06378+00:00",{"id":45,"slug":46,"title":47,"cover_image":48,"image_url":48,"created_at":49,"category":13},"81fa50cf-ee8b-4b76-b017-7dfc45a2dea0","bitcoin-price-page-risk-asset-market-signal-en","Bitcoin’s price page proves the market still treats BTC like a risk a…","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782776869895-brr2.png","2026-06-29T23:47:27.031808+00:00",{"id":51,"slug":52,"title":53,"cover_image":54,"image_url":54,"created_at":55,"category":13},"5408aa94-6f8f-4f20-9629-7c5550859f3b","sora-smash-ultimate-final-dlc-pick-balanced-en","Sora in Smash Ultimate is a strong final DLC pick, not a broken one","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782775073444-djk4.png","2026-06-29T23:17:22.741007+00:00",{"id":57,"slug":58,"title":59,"cover_image":60,"image_url":60,"created_at":61,"category":13},"08cd2ab1-2a2c-4ab6-ab51-4b16a0fed4ab","openclaw-135000-star-saas-security-crisis-en","135,000-star OpenClaw hits SaaS security crisis","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782771534012-jg28.png","2026-06-29T22:17:16.610831+00:00",{"id":63,"slug":64,"title":65,"cover_image":66,"image_url":66,"created_at":67,"category":13},"13701fd7-c4c2-4966-a6e7-db3646d99bd7","anthropic-ipo-965b-valuation-sec-filing-en","Anthropic IPO: $965B valuation and SEC filing","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782770565405-h3hj.png","2026-06-29T22:02:19.831993+00:00",{"id":69,"slug":70,"title":71,"cover_image":72,"image_url":72,"created_at":73,"category":13},"9f3418e2-07ff-4903-a189-6fbe97d079da","hp-openai-frontier-partnership-en","HP and OpenAI expand Frontier partnership","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782766963986-pbe2.png","2026-06-29T21:02:22.652434+00:00",[75,80,85,90,95,100,105,110,115,120],{"id":76,"slug":77,"title":78,"created_at":79},"d35a1bd9-e709-412e-a2df-392df1dc572a","ai-impact-2026-developments-market-en","AI's Impact in 2026: Key Developments and Market Shifts","2026-03-25T16:20:33.205823+00:00",{"id":81,"slug":82,"title":83,"created_at":84},"5ed27921-5fd6-492e-8c59-78393bf37710","trumps-ai-legislative-framework-en","Trump's AI Legislative Framework: What's Inside?","2026-03-25T16:22:20.005325+00:00",{"id":86,"slug":87,"title":88,"created_at":89},"e454a642-f03c-4794-b185-5f651aebbaca","nvidia-gtc-2026-key-highlights-innovations-en","NVIDIA GTC 2026: Key Highlights and Innovations","2026-03-25T16:22:47.882615+00:00",{"id":91,"slug":92,"title":93,"created_at":94},"0ebb5b16-774a-4922-945d-5f2ce1df5a6d","claude-usage-diversifies-learning-curves-en","Claude Usage Diversifies, Learning Curves Emerge","2026-03-25T16:25:50.770376+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"69934e86-2fc5-4280-8223-7b917a48ace8","openclaw-ai-commoditization-concerns-en","OpenClaw's Rise Raises Concerns of AI Model Commoditization","2026-03-25T16:26:30.582047+00:00",{"id":101,"slug":102,"title":103,"created_at":104},"b4b2575b-2ac8-46b2-b90e-ab1d7c060797","google-gemini-ai-rollout-2026-en","Google's Gemini AI Rollout Extended to 2026","2026-03-25T16:28:14.808842+00:00",{"id":106,"slug":107,"title":108,"created_at":109},"6e18bc65-42ae-4ad0-b564-67d7f66b979e","meta-llama4-fabricated-results-scandal-en","Meta's Llama 4 Scandal: Fabricated AI Test Results Unveiled","2026-03-25T16:29:15.482836+00:00",{"id":111,"slug":112,"title":113,"created_at":114},"bf888e9d-08be-4f47-996c-7b24b5ab3500","accenture-mistral-ai-deployment-en","Accenture and Mistral AI Team Up for AI Deployment","2026-03-25T16:31:01.894655+00:00",{"id":116,"slug":117,"title":118,"created_at":119},"5382b536-fad2-49c6-ac85-9eb2bae49f35","mistral-ai-high-stakes-2026-en","Mistral AI: Facing High Stakes in 2026","2026-03-25T16:31:39.941974+00:00",{"id":121,"slug":122,"title":123,"created_at":124},"9da3d2d6-b669-4971-ba1d-17fdb3548ed5","cursors-meteoric-rise-pressures-en","Cursor's Meteoric Rise Faces Industry Pressures","2026-03-25T16:32:21.899217+00:00"]