{"product_id":"gamengen","title":"GameNGen","description":"\u003cstyle\u003e\n.guidaio-v3 { --gv3-bg:#f6f5fb;--gv3-surface:#fff;--gv3-surface-soft:#fbfaff;--gv3-ink:#14132b;--gv3-ink-2:#2b2a4a;--gv3-muted:#6c6a8a;--gv3-line:#ecebf5;--gv3-line-2:#e0deef;--gv3-brand:#5b3df5;--gv3-brand-2:#7c63ff;--gv3-brand-soft:#efeaff;--gv3-brand-ink:#2c1d8a;--gv3-ok:#0f8a5a;--gv3-ok-soft:#e6f7ee;--gv3-warn:#b16800;--gv3-warn-soft:#fff3dc;--gv3-bad:#b3261e;--gv3-bad-soft:#fde7e6;--gv3-radius-lg:22px;--gv3-radius:16px;--gv3-radius-sm:10px;--gv3-shadow-sm:0 1px 2px rgba(20,19,43,.04);--gv3-shadow:0 12px 30px rgba(53,38,124,.07);\n  background:radial-gradient(1200px 600px at 80% -10%,rgba(124,99,255,.18),transparent 60%),radial-gradient(900px 500px at -10% 5%,rgba(91,61,245,.12),transparent 60%),var(--gv3-bg);\n  color:var(--gv3-ink);font-family:Inter,system-ui,-apple-system,\"Segoe UI\",Roboto,sans-serif;font-size:16px;line-height:1.6;-webkit-font-smoothing:antialiased;\n  max-width:1240px;margin:0 auto;padding:24px 16px 64px;border-radius:var(--gv3-radius-lg);}\n.guidaio-v3 *,.guidaio-v3 *::before,.guidaio-v3 *::after{box-sizing:border-box}\n.guidaio-v3 a{color:var(--gv3-brand);text-decoration:none}.guidaio-v3 a:hover{text-decoration:underline}\n.guidaio-v3 img{max-width:100%;display:block}\n.guidaio-v3 h1,.guidaio-v3 h2,.guidaio-v3 h3,.guidaio-v3 h4{font-family:\"Space Grotesk\",Inter,sans-serif;letter-spacing:-.01em;margin:0;color:var(--gv3-ink)}\n.guidaio-v3 svg{display:inline-block;vertical-align:middle}\n.guidaio-v3 .hero{position:relative;background:radial-gradient(700px 320px at 90% 0%,rgba(124,99,255,.18),transparent 70%),linear-gradient(180deg,#fff 0%,#fbfaff 100%);border:1px solid var(--gv3-line);border-radius:var(--gv3-radius-lg);box-shadow:var(--gv3-shadow);padding:32px;overflow:hidden}\n.guidaio-v3 .hero-grid{display:grid;grid-template-columns:minmax(0,1.5fr) minmax(280px,1fr);gap:28px;align-items:start}\n.guidaio-v3 .hero-head{display:flex;gap:16px;align-items:flex-start}\n.guidaio-v3 .logo-tile{width:72px;height:72px;border-radius:18px;border:1px solid var(--gv3-line-2);background:#fff;box-shadow:var(--gv3-shadow-sm);display:grid;place-items:center;overflow:hidden;flex:0 0 auto}\n.guidaio-v3 .logo-tile img{width:56px;height:56px;object-fit:contain}\n.guidaio-v3 .eyebrow{display:inline-flex;align-items:center;gap:6px;color:var(--gv3-brand-ink);background:var(--gv3-brand-soft);padding:6px 12px;border-radius:999px;font-size:12px;font-weight:700;letter-spacing:.04em;text-transform:uppercase}\n.guidaio-v3 h1.title{margin-top:12px;font-size:34px;line-height:1.1;color:var(--gv3-ink)}\n.guidaio-v3 .lede{margin:14px 0 0;font-size:17px;color:var(--gv3-ink-2);max-width:60ch}\n.guidaio-v3 .badge-row,.guidaio-v3 .chip-row{display:flex;flex-wrap:wrap;gap:8px;margin-top:16px}\n.guidaio-v3 .badge{display:inline-flex;align-items:center;gap:6px;padding:6px 11px;border-radius:999px;font-size:12px;font-weight:600;background:#fff;border:1px solid var(--gv3-line-2);color:var(--gv3-ink-2)}\n.guidaio-v3 .badge.ok{background:var(--gv3-ok-soft);color:var(--gv3-ok);border-color:#b8e8cc}\n.guidaio-v3 .badge.warn{background:var(--gv3-warn-soft);color:var(--gv3-warn);border-color:#f1d6a3}\n.guidaio-v3 .badge.bad{background:var(--gv3-bad-soft);color:var(--gv3-bad);border-color:#f5c2c0}\n.guidaio-v3 .badge.brand{background:var(--gv3-brand-soft);color:var(--gv3-brand-ink);border-color:#d6cdfb}\n.guidaio-v3 .tbd-row{margin-top:14px;display:grid;gap:10px}\n.guidaio-v3 .tbd-row .label{font-size:11px;font-weight:700;letter-spacing:.06em;text-transform:uppercase;color:var(--gv3-muted)}\n.guidaio-v3 .chip{display:inline-flex;align-items:center;gap:6px;padding:5px 11px;border-radius:999px;font-size:13px;border:1px solid var(--gv3-line-2);background:#fff;color:var(--gv3-ink-2)}\n.guidaio-v3 .chip.cat{background:#eef2ff;color:#3730a3;border-color:#c7d2fe}\n.guidaio-v3 .chip.dom{background:#ecfeff;color:#0e7490;border-color:#a5f3fc}\n.guidaio-v3 .chip.prof{background:#fdf4ff;color:#86198f;border-color:#f5d0fe}\n.guidaio-v3 .cta-row{display:flex;gap:10px;flex-wrap:wrap;margin-top:22px}\n.guidaio-v3 .btn{display:inline-flex;align-items:center;gap:8px;padding:11px 16px;border-radius:12px;font-weight:700;border:1px solid transparent;cursor:pointer;font-size:14px}\n.guidaio-v3 .btn:hover{text-decoration:none}\n.guidaio-v3 .btn-primary{background:linear-gradient(135deg,var(--gv3-brand) 0%,var(--gv3-brand-2) 100%);color:#fff;box-shadow:0 14px 28px rgba(91,61,245,.28)}\n.guidaio-v3 .btn-secondary{background:#fff;color:var(--gv3-brand-ink);border-color:#d6cdfb}\n.guidaio-v3 .btn-ghost{background:transparent;color:var(--gv3-ink-2);border-color:var(--gv3-line-2)}\n.guidaio-v3 .affiliate-note{margin-top:10px;font-size:12px;color:var(--gv3-muted)}\n.guidaio-v3 .snapshot{background:#fff;border:1px solid var(--gv3-line);border-radius:var(--gv3-radius-lg);padding:20px;box-shadow:var(--gv3-shadow)}\n.guidaio-v3 .snapshot h3{font-size:13px;text-transform:uppercase;letter-spacing:.08em;color:var(--gv3-brand-ink);margin-bottom:14px}\n.guidaio-v3 .snapshot dl{margin:0;display:grid;grid-template-columns:1fr;gap:12px}\n.guidaio-v3 .snapshot .row{display:grid;grid-template-columns:1fr auto;gap:8px;padding-bottom:12px;border-bottom:1px dashed var(--gv3-line-2)}\n.guidaio-v3 .snapshot .row:last-child{border-bottom:0;padding-bottom:0}\n.guidaio-v3 .snapshot dt{color:var(--gv3-muted);font-size:13px;margin:0}\n.guidaio-v3 .snapshot dd{margin:0;text-align:right;font-weight:700;color:var(--gv3-ink);font-size:13px}\n.guidaio-v3 .snapshot dd small{display:block;font-weight:500;color:var(--gv3-muted);font-size:11px;text-align:right}\n.guidaio-v3 .price-big{font-family:\"Space Grotesk\";font-size:20px;line-height:1}\n.guidaio-v3 .toc{margin:20px 0 24px;padding:10px;background:rgba(255,255,255,.85);border:1px solid var(--gv3-line);border-radius:22px;box-shadow:var(--gv3-shadow-sm);display:flex;flex-wrap:wrap;gap:8px;overflow-x:auto}\n.guidaio-v3 .toc a{padding:8px 14px;border-radius:999px;font-size:13px;font-weight:600;color:var(--gv3-brand-ink);background:#fff;white-space:nowrap;border:1px solid var(--gv3-line-2);box-shadow:var(--gv3-shadow-sm);transition:transform .18s ease,box-shadow .18s ease,background .18s ease,color .18s ease,border-color .18s ease}\n.guidaio-v3 .toc a:hover{background:linear-gradient(135deg,var(--gv3-brand) 0%,var(--gv3-brand-2) 100%);color:#fff;border-color:transparent;transform:translateY(-2px) scale(1.04);box-shadow:0 8px 18px rgba(91,61,245,.25);text-decoration:none}\n.guidaio-v3 .justify{text-align:justify;text-justify:inter-word;hyphens:auto}\n.guidaio-v3 a.chip,.guidaio-v3 a.chip:hover{text-decoration:none}\n.guidaio-v3 a.chip{transition:transform .15s ease,box-shadow .15s ease,filter .15s ease;cursor:pointer}\n.guidaio-v3 a.chip:hover{transform:translateY(-1px);box-shadow:0 4px 10px rgba(53,38,124,.12);filter:brightness(.96)}\n.guidaio-v3 section.block{background:var(--gv3-surface);border:1px solid var(--gv3-line);border-radius:var(--gv3-radius-lg);padding:32px;margin-top:22px;box-shadow:var(--gv3-shadow-sm);scroll-margin-top:90px}\n.guidaio-v3 section.block.soft{background:var(--gv3-surface-soft)}\n.guidaio-v3 section.block .head{margin-bottom:20px}\n.guidaio-v3 section.block .eyebrow + h2{margin-top:10px}\n.guidaio-v3 section.block h2{font-size:24px;color:var(--gv3-ink)}\n.guidaio-v3 section.block .lead{margin-top:8px;color:var(--gv3-muted);max-width:70ch}\n.guidaio-v3 .grid-2{display:grid;grid-template-columns:repeat(2,minmax(0,1fr));gap:16px}\n.guidaio-v3 .grid-3{display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:14px}\n.guidaio-v3 .rich p{margin:0 0 14px}.guidaio-v3 .rich p:last-child{margin:0}\n.guidaio-v3 ul.brand-list{list-style:none;padding:0;margin:0;display:grid;gap:10px}\n.guidaio-v3 ul.brand-list li{position:relative;padding-left:26px;color:var(--gv3-ink-2)}\n.guidaio-v3 ul.brand-list li::before{content:\"\";position:absolute;left:0;top:9px;width:14px;height:14px;border-radius:999px;background:var(--gv3-brand-soft);box-shadow:inset 0 0 0 4px var(--gv3-brand)}\n.guidaio-v3 ol.steps{list-style:none;counter-reset:s;padding:0;margin:0;display:grid;gap:14px}\n.guidaio-v3 ol.steps li{counter-increment:s;position:relative;padding-left:42px;color:var(--gv3-ink-2);min-height:30px}\n.guidaio-v3 ol.steps li::before{content:counter(s);position:absolute;left:0;top:0;width:30px;height:30px;border-radius:999px;background:linear-gradient(135deg,var(--gv3-brand) 0%,var(--gv3-brand-2) 100%);color:#fff;display:grid;place-items:center;font-weight:700;font-size:13px;box-shadow:0 6px 14px rgba(91,61,245,.25)}\n.guidaio-v3 .feature-card{border:1px solid var(--gv3-line);border-radius:var(--gv3-radius);padding:16px;background:#fff;display:grid;gap:8px}\n.guidaio-v3 .feature-card .icbox{width:36px;height:36px;border-radius:12px;background:var(--gv3-brand-soft);color:var(--gv3-brand-ink);display:grid;place-items:center}\n.guidaio-v3 .feature-card .ftitle{font-weight:700;color:var(--gv3-ink);font-size:15px}\n.guidaio-v3 .feature-card .fbody{color:var(--gv3-muted);font-size:13px}\n.guidaio-v3 .pc{border-radius:var(--gv3-radius);padding:20px;border:1px solid var(--gv3-line)}\n.guidaio-v3 .pc.pros{background:var(--gv3-ok-soft);border-color:#b8e8cc}\n.guidaio-v3 .pc.cons{background:var(--gv3-bad-soft);border-color:#f5c2c0}\n.guidaio-v3 .pc h3{font-size:17px;display:flex;align-items:center;gap:8px;margin-bottom:12px}\n.guidaio-v3 .pc.pros h3{color:var(--gv3-ok)}.guidaio-v3 .pc.cons h3{color:var(--gv3-bad)}\n.guidaio-v3 .pc ul{list-style:none;padding:0;margin:0;display:grid;gap:8px}\n.guidaio-v3 .pc li{display:flex;gap:8px;align-items:flex-start;color:var(--gv3-ink-2);font-size:14px}\n.guidaio-v3 .pc li svg{flex:0 0 auto;margin-top:3px}\n.guidaio-v3 .pc.pros li svg{color:var(--gv3-ok)}.guidaio-v3 .pc.cons li svg{color:var(--gv3-bad)}\n.guidaio-v3 .bestnot{display:grid;grid-template-columns:1fr 1fr;gap:16px}\n.guidaio-v3 .bestnot .card{border:1px solid var(--gv3-line);border-radius:var(--gv3-radius);padding:20px;background:#fff}\n.guidaio-v3 .bestnot h3{display:flex;align-items:center;gap:8px;font-size:17px;margin-bottom:12px}\n.guidaio-v3 .bestnot .for h3{color:var(--gv3-ok)}.guidaio-v3 .bestnot .not h3{color:var(--gv3-warn)}\n.guidaio-v3 .plan-grid{display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:14px}\n.guidaio-v3 .plan{border:1px solid var(--gv3-line);border-radius:var(--gv3-radius-lg);padding:22px;background:#fff;display:grid;gap:10px;position:relative}\n.guidaio-v3 .plan.featured{border-color:#c7baff;background:linear-gradient(180deg,#fff 0%,#f7f3ff 100%);box-shadow:var(--gv3-shadow)}\n.guidaio-v3 .plan .pill{position:absolute;top:-12px;right:14px;background:linear-gradient(135deg,var(--gv3-brand) 0%,var(--gv3-brand-2) 100%);color:#fff;font-size:11px;font-weight:700;letter-spacing:.06em;padding:5px 9px;border-radius:999px;text-transform:uppercase}\n.guidaio-v3 .plan .pname{font-family:\"Space Grotesk\";font-weight:700;font-size:17px}\n.guidaio-v3 .plan .pprice{font-family:\"Space Grotesk\";font-size:24px;font-weight:700}\n.guidaio-v3 .plan .pprice small{font-size:13px;font-weight:500;color:var(--gv3-muted)}\n.guidaio-v3 .plan ul{list-style:none;padding:0;margin:6px 0 0;display:grid;gap:8px}\n.guidaio-v3 .plan ul li{display:flex;gap:8px;align-items:flex-start;color:var(--gv3-ink-2);font-size:14px}\n.guidaio-v3 .plan ul li svg{color:var(--gv3-brand);flex:0 0 auto;margin-top:3px}\n.guidaio-v3 .pricing-extras{margin-top:16px;display:grid;gap:10px}\n.guidaio-v3 .offer{display:flex;gap:10px;align-items:flex-start;padding:12px 14px;border-radius:12px;background:#fff7e6;border:1px solid #f1d6a3;color:var(--gv3-warn);font-size:14px}\n.guidaio-v3 .offer svg{flex:0 0 auto;margin-top:2px}\n.guidaio-v3 .trust-grid{display:grid;grid-template-columns:2fr 1fr;gap:16px}\n.guidaio-v3 .trust-grid .col{display:grid;gap:14px}\n.guidaio-v3 .trust-card{background:#fff;border:1px solid var(--gv3-line);border-radius:var(--gv3-radius);padding:20px}\n.guidaio-v3 .trust-card h3{font-size:15px;margin-bottom:10px}\n.guidaio-v3 .kv{display:grid;grid-template-columns:180px 1fr;gap:12px;padding:10px 0;border-top:1px solid var(--gv3-line)}\n.guidaio-v3 .kv:first-child{border-top:0;padding-top:0}\n.guidaio-v3 .kv .k{color:var(--gv3-muted);font-size:13px}\n.guidaio-v3 .kv .v{color:var(--gv3-ink);font-size:14px}\n.guidaio-v3 .availability{display:grid;grid-template-columns:1fr 1fr;gap:16px}\n.guidaio-v3 .availability .card{border:1px solid var(--gv3-line);border-radius:var(--gv3-radius);background:#fff;padding:18px}\n.guidaio-v3 .flag-row{display:flex;flex-wrap:wrap;gap:8px}\n.guidaio-v3 .flag{display:inline-flex;align-items:center;gap:5px;padding:5px 10px;border-radius:999px;background:#fff;border:1px solid var(--gv3-line-2);font-size:13px;color:var(--gv3-ink-2)}\n.guidaio-v3 section.block.watch{background:#fffaf0;border:1px solid #f1d6a3}\n.guidaio-v3 section.block.watch .head h2{color:var(--gv3-warn)}\n.guidaio-v3 .setup-grid{display:grid;grid-template-columns:1fr 1fr;gap:16px}\n.guidaio-v3 .setup-card{background:#fff;border:1px solid var(--gv3-line);border-radius:var(--gv3-radius);padding:20px}\n.guidaio-v3 .integrations{display:flex;flex-wrap:wrap;gap:8px}\n.guidaio-v3 .integration{display:inline-flex;align-items:center;gap:6px;padding:7px 11px;border-radius:12px;background:#fff;border:1px solid var(--gv3-line-2);color:var(--gv3-ink-2);font-size:13px;font-weight:600}\n.guidaio-v3 .company-grid{display:grid;grid-template-columns:1.4fr 1fr;gap:16px}\n.guidaio-v3 .company-card{border:1px solid var(--gv3-line);border-radius:var(--gv3-radius);background:#fff;padding:20px}\n.guidaio-v3 .timeline{display:grid;gap:8px;margin:8px 0 0}\n.guidaio-v3 .timeline .ev{padding:10px 14px;border-radius:12px;background:var(--gv3-brand-soft);color:var(--gv3-brand-ink);font-weight:600;font-size:14px;display:flex;align-items:center;gap:8px}\n.guidaio-v3 .social-row{display:flex;gap:8px;flex-wrap:wrap}\n.guidaio-v3 .social{width:36px;height:36px;border-radius:12px;border:1px solid var(--gv3-line-2);background:#fff;display:grid;place-items:center;color:var(--gv3-ink-2)}\n.guidaio-v3 .social:hover{background:var(--gv3-brand-soft);color:var(--gv3-brand-ink);border-color:#d6cdfb}\n.guidaio-v3 .res-grid{display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:10px}\n.guidaio-v3 .res{display:flex;gap:10px;align-items:center;padding:10px 12px;border-radius:12px;border:1px solid var(--gv3-line);background:#fff;color:var(--gv3-ink-2);font-size:13px}\n.guidaio-v3 .res:hover{border-color:#d6cdfb;background:#fbfaff;text-decoration:none}\n.guidaio-v3 .res svg{color:var(--gv3-brand);flex:0 0 auto}\n.guidaio-v3 .res .url{color:var(--gv3-muted);font-size:11px;word-break:break-all}\n.guidaio-v3 .res .res-text{display:grid}\n.guidaio-v3 .alts{display:flex;flex-wrap:wrap;gap:8px}\n.guidaio-v3 .alt{display:inline-flex;align-items:center;gap:8px;padding:9px 13px;border-radius:14px;background:#fff;border:1px solid var(--gv3-line-2);font-weight:600;color:var(--gv3-ink-2);font-size:14px}\n.guidaio-v3 .alt .av{width:24px;height:24px;border-radius:8px;background:linear-gradient(135deg,#c7baff 0%,#efeaff 100%);display:grid;place-items:center;font-size:12px;font-weight:700;color:var(--gv3-brand-ink)}\n.guidaio-v3 .faq{display:grid;gap:10px}\n.guidaio-v3 .faq details{background:#fff;border:1px solid var(--gv3-line);border-radius:var(--gv3-radius);padding:14px 16px}\n.guidaio-v3 .faq summary{cursor:pointer;list-style:none;font-weight:700;color:var(--gv3-ink);display:flex;justify-content:space-between;align-items:center;gap:10px;font-size:15px}\n.guidaio-v3 .faq summary::-webkit-details-marker{display:none}\n.guidaio-v3 .faq summary svg{color:var(--gv3-brand)}\n.guidaio-v3 .faq .ans{color:var(--gv3-muted);margin-top:10px;font-size:14px}\n.guidaio-v3 section.block.conclusion{background:radial-gradient(700px 320px at 0% 0%,rgba(124,99,255,.18),transparent 60%),linear-gradient(135deg,#1c163a 0%,#2c1d8a 100%);color:#fff;border-color:#2c1d8a}\n.guidaio-v3 section.block.conclusion .head h2{color:#fff}\n.guidaio-v3 section.block.conclusion .lead,.guidaio-v3 section.block.conclusion .rich p{color:rgba(255,255,255,.88)}\n.guidaio-v3 section.block.conclusion .rich strong,.guidaio-v3 section.block.conclusion .rich b{color:#fff}\n.guidaio-v3 section.block.conclusion .eyebrow{background:rgba(255,255,255,.14);color:#fff}\n.guidaio-v3 section.block.conclusion .btn-secondary{background:rgba(255,255,255,.14);color:#fff;border-color:rgba(255,255,255,.25)}\n@media (max-width:1080px){.guidaio-v3 .hero-grid,.guidaio-v3 .trust-grid,.guidaio-v3 .company-grid,.guidaio-v3 .setup-grid,.guidaio-v3 .availability{grid-template-columns:1fr}.guidaio-v3 .grid-3,.guidaio-v3 .res-grid,.guidaio-v3 .plan-grid{grid-template-columns:repeat(2,minmax(0,1fr))}}\n@media (max-width:700px){.guidaio-v3{padding:12px 8px 48px}.guidaio-v3 .hero{padding:20px}.guidaio-v3 section.block{padding:20px}.guidaio-v3 .grid-2,.guidaio-v3 .grid-3,.guidaio-v3 .res-grid,.guidaio-v3 .plan-grid,.guidaio-v3 .bestnot,.guidaio-v3 .availability{grid-template-columns:1fr}.guidaio-v3 h1.title{font-size:28px}}\n\u003c\/style\u003e\n\u003cdiv class=\"guidaio-v3\" itemscope itemtype=\"https:\/\/schema.org\/SoftwareApplication\"\u003e\n\u003cmeta itemprop=\"name\" content=\"GameNGen: Neural Game Engine Simulating DOOM in Real Time\"\u003e\n\u003cmeta itemprop=\"description\" content=\"GameNGen simulates DOOM in real time with a diffusion model, over 20 fps on one TPU. Google research page, arXiv\/ICLR 2025 paper: no code, no demo, no price.\"\u003e\n\u003cdiv class=\"hero\"\u003e\u003cdiv class=\"hero-grid\"\u003e\n\u003cdiv\u003e\n\u003cdiv class=\"hero-head\"\u003e\n\u003cdiv class=\"logo-tile\"\u003e\u003cimg src=\"https:\/\/gamengen.github.io\/static\/images\/embed_image.png\" alt=\"GameNGen logo\" itemprop=\"image\" loading=\"lazy\"\u003e\u003c\/div\u003e\n\u003cdiv class=\"name-block\"\u003e\n\u003cspan class=\"eyebrow\"\u003e\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cpath d=\"m12 3-1.9 5.8a2 2 0 0 1-1.3 1.3L3 12l5.8 1.9a2 2 0 0 1 1.3 1.3L12 21l1.9-5.8a2 2 0 0 1 1.3-1.3L21 12l-5.8-1.9a2 2 0 0 1-1.3-1.3Z\"\u003e\u003c\/path\u003e\u003c\/svg\u003e Academic Research · Gaming Tools\u003c\/span\u003e\u003ch1 class=\"title\" itemprop=\"name\"\u003eGameNGen\u003c\/h1\u003e\n\u003cp class=\"lede justify\" itemprop=\"description\"\u003eGameNGen is a Google \u003cb\u003eresearch result\u003c\/b\u003e, not a usable product: a neural model that interactively simulates \u003ci\u003eDOOM\u003c\/i\u003e (1993) at over 20 frames per second on a single TPU. Published at \u003cb\u003eICLR 2025\u003c\/b\u003e; no code, no demo.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"badge-row\"\u003e\n\u003cspan class=\"badge ok\"\u003e\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003ccircle cx=\"12\" cy=\"12\" r=\"10\"\u003e\u003c\/circle\u003e\u003cpath d=\"m9 12 2 2 4-4\"\u003e\u003c\/path\u003e\u003c\/svg\u003e Active\u003c\/span\u003e\u003cspan class=\"badge\"\u003e\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cpath d=\"M6.3 20.3a2.4 2.4 0 0 0 3.4 0L12 18l-6-6-2.3 2.3a2.4 2.4 0 0 0 0 3.4Z\"\u003e\u003c\/path\u003e\u003cpath d=\"m2 22 3-3\"\u003e\u003c\/path\u003e\u003cpath d=\"M7.5 13.5 10 11\"\u003e\u003c\/path\u003e\u003cpath d=\"M10.5 16.5 13 14\"\u003e\u003c\/path\u003e\u003cpath d=\"m18 3-4 4h6l-4 4\"\u003e\u003c\/path\u003e\u003c\/svg\u003e No public API\u003c\/span\u003e\u003cspan class=\"badge brand\"\u003e\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003ccircle cx=\"12\" cy=\"12\" r=\"10\"\u003e\u003c\/circle\u003e\u003cpath d=\"m9 12 2 2 4-4\"\u003e\u003c\/path\u003e\u003c\/svg\u003e Verified by Guidaio\u003c\/span\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"tbd-row\"\u003e\n\u003cdiv\u003e\n\u003cdiv class=\"label\"\u003eCategories\u003c\/div\u003e\n\u003cdiv class=\"chip-row\"\u003e\n\u003ca class=\"chip cat\" href=\"\/collections\/academic-research\"\u003eAcademic Research\u003c\/a\u003e\u003ca class=\"chip cat\" href=\"\/collections\/gaming-tools\"\u003eGaming Tools\u003c\/a\u003e\u003ca class=\"chip cat\" href=\"\/collections\/video-generators\"\u003eVideo Generators\u003c\/a\u003e\u003ca class=\"chip cat\" href=\"\/collections\/media-models\"\u003eMedia Models\u003c\/a\u003e\u003ca class=\"chip cat\" href=\"\/collections\/fun-experiments\"\u003eFun Experiments\u003c\/a\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\n\u003cdiv\u003e\n\u003cdiv class=\"label\"\u003eTarget domains\u003c\/div\u003e\n\u003cdiv class=\"chip-row\"\u003e\n\u003ca class=\"chip dom\" href=\"\/collections\/data-ai\"\u003eData Ai\u003c\/a\u003e\u003ca class=\"chip dom\" href=\"\/collections\/science-r-d\"\u003eScience R D\u003c\/a\u003e\u003ca class=\"chip dom\" href=\"\/collections\/education-academia\"\u003eEducation Academia\u003c\/a\u003e\u003ca class=\"chip dom\" href=\"\/collections\/it-software\"\u003eIt Software\u003c\/a\u003e\u003ca class=\"chip dom\" href=\"\/collections\/arts-entertainment\"\u003eArts Entertainment\u003c\/a\u003e\u003ca class=\"chip dom\" href=\"\/collections\/film-tv-video\"\u003eFilm Tv Video\u003c\/a\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"cta-row\"\u003e\u003ca class=\"btn btn-primary\" href=\"https:\/\/gamengen.github.io\/\" target=\"_blank\" rel=\"noopener\"\u003e\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"16\" height=\"16\"\u003e\u003cpath d=\"M15 3h6v6\"\u003e\u003c\/path\u003e\u003cpath d=\"M10 14 21 3\"\u003e\u003c\/path\u003e\u003cpath d=\"M18 13v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6\"\u003e\u003c\/path\u003e\u003c\/svg\u003e Visit the project page\u003c\/a\u003e\u003c\/div\u003e\n\u003c\/div\u003e\n\u003caside class=\"snapshot\"\u003e\u003ch3\u003eSnapshot\u003c\/h3\u003e\n\u003cdl\u003e\n\u003cdiv class=\"row\"\u003e\n\u003cdt\u003eAPI available\u003c\/dt\u003e\n\u003cdd\u003eNo\u003c\/dd\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"row\"\u003e\n\u003cdt\u003eTool status\u003c\/dt\u003e\n\u003cdd\u003eActive\u003c\/dd\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"row\"\u003e\n\u003cdt\u003eCountry of origin\u003c\/dt\u003e\n\u003cdd\u003e🇺🇸 United States\u003c\/dd\u003e\n\u003c\/div\u003e\n\u003c\/dl\u003e\u003c\/aside\u003e\n\u003c\/div\u003e\u003c\/div\u003e\n\u003cnav class=\"toc\" aria-label=\"Page sections\"\u003e\u003ca href=\"#what\"\u003eWhat is it\u003c\/a\u003e\u003ca href=\"#bestnot\"\u003eWhen to use GameNGen \/ When not to\u003c\/a\u003e\u003ca href=\"#how\"\u003eHow to use\u003c\/a\u003e\u003ca href=\"#proscons\"\u003ePros \u0026amp; Cons\u003c\/a\u003e\u003ca href=\"#pricing\"\u003ePricing\u003c\/a\u003e\u003ca href=\"#trust\"\u003eTrust \u0026amp; Privacy\u003c\/a\u003e\u003ca href=\"#watch\"\u003eWatch-outs\u003c\/a\u003e\u003ca href=\"#setup\"\u003eSetup \u0026amp; Integrations\u003c\/a\u003e\u003ca href=\"#company\"\u003eCompany\u003c\/a\u003e\u003ca href=\"#resources\"\u003eResources\u003c\/a\u003e\u003ca href=\"#faq\"\u003eFAQ\u003c\/a\u003e\u003ca href=\"#conclusion\"\u003eConclusion\u003c\/a\u003e\u003c\/nav\u003e\n\u003csection class=\"block\" id=\"what\"\u003e\u003cdiv class=\"head\"\u003e\n\u003cspan class=\"eyebrow\"\u003eOverview\u003c\/span\u003e\u003ch2\u003eWhat is GameNGen?\u003c\/h2\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"grid-2\"\u003e\n\u003cdiv class=\"rich justify\"\u003e\n\u003cp\u003eGameNGen is a research system presented in the paper \u003ci\u003eDiffusion Models Are Real-Time Game Engines\u003c\/i\u003e (arXiv 2408.14837, accepted at ICLR 2025). Its central claim is to be \u003cb\u003ethe first game engine powered entirely by a neural model\u003c\/b\u003e, one that also enables real-time interaction with a complex environment over long trajectories and at high quality. The object of the demonstration is DOOM (1993), simulated end to end by the model, with no classical game engine running underneath.\u003c\/p\u003e\n\u003cp\u003eThe reported numbers are concrete. The system runs at more than 20 frames per second on a single TPU and stays stable over sessions of several minutes of auto-regressive generation. Next-frame prediction reaches a PSNR of 29.4, which the authors compare to lossy JPEG compression. In a human evaluation, raters do only slightly better than chance when asked to tell short clips of the real game from clips of the simulation.\u003c\/p\u003e\n\u003cp\u003eTraining runs in two phases. First, a reinforcement learning agent learns to play DOOM, and its episodes of actions and observations are recorded, since human gameplay could not be collected at that scale. Second, a diffusion model is trained on those recordings to produce the next frame, conditioned jointly on the sequence of past frames and past actions. The base model is \u003cb\u003eStable Diffusion v1.4\u003c\/b\u003e, repurposed rather than built from scratch.\u003c\/p\u003e\n\u003cp\u003eTwo engineering choices carry the result. To limit auto-regressive drift, context frames are corrupted with Gaussian noise added to the encoded frames during training, teaching the network to correct information sampled in previous frames; the authors judge this critical for visual stability over long periods. And because the pretrained autoencoder compresses 8x8 pixel patches into four latent channels, producing artefacts on fine detail and above all on the bottom HUD bar, the decoder alone is retrained with an MSE loss against the target frame pixels; the revised paper reports better fidelity on small details and on text.\u003c\/p\u003e\n\u003cp\u003eThe work is signed by Dani Valevski and Yaniv Leviathan (Google Research), Moab Arar (Tel Aviv University, work done at Google Research) and Shlomi Fruchter (Google DeepMind), with equal contribution. \u003cb\u003eWhat is published is a result, not software\u003c\/b\u003e: no code, no weights, no checkpoint and no playable demo have been released.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003cdiv\u003e\n\u003ch3 style=\"font-size:14px;margin-bottom:12px;color:var(--gv3-brand-ink);text-transform:uppercase;letter-spacing:.08em\"\u003eWhat it does\u003c\/h3\u003e\n\u003cul class=\"brand-list\"\u003e\n\u003cli\u003eSimulates DOOM interactively and in real time from a single neural model\u003c\/li\u003e\n\u003cli\u003ePredicts the next frame conditioned on the sequence of past frames and past player actions\u003c\/li\u003e\n\u003cli\u003eRuns at more than 20 frames per second on a single TPU\u003c\/li\u003e\n\u003cli\u003eStays stable over long auto-regressive trajectories, across sessions of several minutes\u003c\/li\u003e\n\u003cli\u003eReaches a PSNR of 29.4 on next-frame prediction\u003c\/li\u003e\n\u003cli\u003eProduces frames that human raters barely distinguish from the real game on short clips\u003c\/li\u003e\n\u003cli\u003eLearns the environment from recordings of a reinforcement learning agent, with no classical game engine underneath\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\u003c\/section\u003e\n\u003csection class=\"block soft\" id=\"bestnot\"\u003e\u003cdiv class=\"head\"\u003e\n\u003cspan class=\"eyebrow\"\u003eAudience\u003c\/span\u003e\u003ch2\u003eWhen to use GameNGen \/ When not to\u003c\/h2\u003e\n\u003cp class=\"lead\"\u003eA quick filter to help you decide if GameNGen is the right fit.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"bestnot\"\u003e\n\u003cdiv class=\"card for\"\u003e\n\u003ch3\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"16\" height=\"16\"\u003e\u003cpath d=\"M7 10v12\"\u003e\u003c\/path\u003e\u003cpath d=\"M15 5.88 14 10h5.83a2 2 0 0 1 1.92 2.56l-2.33 8A2 2 0 0 1 17.5 22H4a2 2 0 0 1-2-2v-8a2 2 0 0 1 2-2h2.76a2 2 0 0 0 1.79-1.11L12 2a3.13 3.13 0 0 1 3 3.88Z\"\u003e\u003c\/path\u003e\u003c\/svg\u003e When to use GameNGen\u003c\/h3\u003e\n\u003cul class=\"brand-list\"\u003e\n\u003cli\u003eGenerative AI researchers working on diffusion models and world models\u003c\/li\u003e\n\u003cli\u003eComputer vision researchers studying next-frame prediction, PSNR and autoencoder artefacts\u003c\/li\u003e\n\u003cli\u003eML engineers concerned with real-time inference budgets, since the system runs above 20 fps on a single TPU\u003c\/li\u003e\n\u003cli\u003eGame developers, studios and technical product managers watching neural generation of interactive environments\u003c\/li\u003e\n\u003cli\u003eML lecturers and students who need a concrete, documented case study of action-conditioned diffusion\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"card not\"\u003e\n\u003ch3\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"16\" height=\"16\"\u003e\u003cpath d=\"M17 14V2\"\u003e\u003c\/path\u003e\u003cpath d=\"M9 18.12 10 14H4.17a2 2 0 0 1-1.92-2.56l2.33-8A2 2 0 0 1 6.5 2H20a2 2 0 0 1 2 2v8a2 2 0 0 1-2 2h-2.76a2 2 0 0 0-1.79 1.11L12 22a3.13 3.13 0 0 1-3-3.88Z\"\u003e\u003c\/path\u003e\u003c\/svg\u003e When not to use GameNGen\u003c\/h3\u003e\n\u003cul class=\"brand-list\"\u003e\n\u003cli\u003ePlayers hoping to play: there is no playable demo and no binary, only recorded videos\u003c\/li\u003e\n\u003cli\u003eDevelopers looking for code or weights: no release exists, and the GameNGen GitHub organisation hosts only the website repository\u003c\/li\u003e\n\u003cli\u003eTeams wanting to embed a game engine in production: no API, no SDK and no product licence\u003c\/li\u003e\n\u003cli\u003eBuyers comparing commercial tools: no pricing, no offer and no support channel\u003c\/li\u003e\n\u003cli\u003eNon-technical readers and compliance officers: the page assumes you will read an ML paper, and publishes no terms, privacy policy or DPA\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\u003c\/section\u003e\n\u003csection class=\"block soft\" id=\"how\"\u003e\u003cdiv class=\"head\"\u003e\n\u003cspan class=\"eyebrow\"\u003eGet started\u003c\/span\u003e\u003ch2\u003eHow to use GameNGen\u003c\/h2\u003e\n\u003cp class=\"lead\"\u003eA typical end-to-end flow, from setup to results.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003col class=\"steps\"\u003e\n\u003cli\u003eSet expectations first: the only available use is to read, watch and cite, since there is no account, no download, no API and no playable demo\u003c\/li\u003e\n\u003cli\u003eOpen the project page and read the abstract, which states the claim and the headline figures\u003c\/li\u003e\n\u003cli\u003eWatch the Full Gameplay Videos section, which shows real-time recordings of DOOM levels e1m1, e1m3, e1m5, e1m9 and e2m2 simulated by the model\u003c\/li\u003e\n\u003cli\u003eRead the Architecture section and its diagram: data collection by a reinforcement learning agent, diffusion model training, latent decoder fine-tuning\u003c\/li\u003e\n\u003cli\u003eOpen the paper on arXiv (2408.14837), as an abstract page or as a PDF, for the full method and the complete evaluation\u003c\/li\u003e\n\u003cli\u003eCopy the BibTeX entry at the bottom of the project page to cite the work, under the key valevski2024diffusionmodelsrealtimegame\u003c\/li\u003e\n\u003cli\u003eAccept that any hands-on work means a full reimplementation from the paper, because nothing is provided\u003c\/li\u003e\n\u003cli\u003eScope that reimplementation realistically: a reinforcement learning agent on DOOM, fine-tuning of Stable Diffusion v1.4, and real-time inference on TPU hardware\u003c\/li\u003e\n\u003c\/ol\u003e\u003c\/section\u003e\n\u003csection class=\"block\" id=\"proscons\"\u003e\u003cdiv class=\"head\"\u003e\n\u003cspan class=\"eyebrow\"\u003eQuick read\u003c\/span\u003e\u003ch2\u003ePros \u0026amp; Cons\u003c\/h2\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"grid-2\"\u003e\n\u003cdiv class=\"pc pros\"\u003e\n\u003ch3\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003ccircle cx=\"12\" cy=\"12\" r=\"10\"\u003e\u003c\/circle\u003e\u003cpath d=\"m9 12 2 2 4-4\"\u003e\u003c\/path\u003e\u003c\/svg\u003e Pros\u003c\/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cpolyline points=\"20 6 9 17 4 12\"\u003e\u003c\/polyline\u003e\u003c\/svg\u003e \u003cspan\u003eLandmark result: the first fully neural game engine running in real time\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cpolyline points=\"20 6 9 17 4 12\"\u003e\u003c\/polyline\u003e\u003c\/svg\u003e \u003cspan\u003ePublished, checkable figures: over 20 fps, PSNR 29.4 and a human rater study\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cpolyline points=\"20 6 9 17 4 12\"\u003e\u003c\/polyline\u003e\u003c\/svg\u003e \u003cspan\u003ePeer-reviewed work, accepted at ICLR 2025\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cpolyline points=\"20 6 9 17 4 12\"\u003e\u003c\/polyline\u003e\u003c\/svg\u003e \u003cspan\u003eMethod described in enough detail to be discussed: two phases, a named base model, an explicit anti-drift technique\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cpolyline points=\"20 6 9 17 4 12\"\u003e\u003c\/polyline\u003e\u003c\/svg\u003e \u003cspan\u003eFree to consult, with no account, no form to fill in and no data-capture wall\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cpolyline points=\"20 6 9 17 4 12\"\u003e\u003c\/polyline\u003e\u003c\/svg\u003e \u003cspan\u003eGameplay videos directly viewable across several levels (e1m1, e1m3, e1m5, e1m9, e2m2)\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cpolyline points=\"20 6 9 17 4 12\"\u003e\u003c\/polyline\u003e\u003c\/svg\u003e \u003cspan\u003eBibTeX entry provided, so citation is immediate\u003c\/span\u003e\n\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"pc cons\"\u003e\n\u003ch3\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"16\" height=\"16\"\u003e\u003ccircle cx=\"12\" cy=\"12\" r=\"10\"\u003e\u003c\/circle\u003e\u003cline x1=\"4.93\" y1=\"4.93\" x2=\"19.07\" y2=\"19.07\"\u003e\u003c\/line\u003e\u003c\/svg\u003e Cons\u003c\/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"18\" y1=\"6\" x2=\"6\" y2=\"18\"\u003e\u003c\/line\u003e\u003cline x1=\"6\" y1=\"6\" x2=\"18\" y2=\"18\"\u003e\u003c\/line\u003e\u003c\/svg\u003e \u003cspan\u003eNo code, no weights and no checkpoint released, which makes the work non-reproducible in practice\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"18\" y1=\"6\" x2=\"6\" y2=\"18\"\u003e\u003c\/line\u003e\u003cline x1=\"6\" y1=\"6\" x2=\"18\" y2=\"18\"\u003e\u003c\/line\u003e\u003c\/svg\u003e \u003cspan\u003eNo playable demo: the visitor watches recordings and cannot test anything\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"18\" y1=\"6\" x2=\"6\" y2=\"18\"\u003e\u003c\/line\u003e\u003cline x1=\"6\" y1=\"6\" x2=\"18\" y2=\"18\"\u003e\u003c\/line\u003e\u003c\/svg\u003e \u003cspan\u003eA single game demonstrated (DOOM, 1993), with no evidence of generalisation to other environments on the page\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"18\" y1=\"6\" x2=\"6\" y2=\"18\"\u003e\u003c\/line\u003e\u003cline x1=\"6\" y1=\"6\" x2=\"18\" y2=\"18\"\u003e\u003c\/line\u003e\u003c\/svg\u003e \u003cspan\u003eHardware requirement: a TPU for real-time inference, out of reach for most readers\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"18\" y1=\"6\" x2=\"6\" y2=\"18\"\u003e\u003c\/line\u003e\u003cline x1=\"6\" y1=\"6\" x2=\"18\" y2=\"18\"\u003e\u003c\/line\u003e\u003c\/svg\u003e \u003cspan\u003eNo legal document at all (terms, privacy policy, legal notice) while the page loads third-party trackers\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"18\" y1=\"6\" x2=\"6\" y2=\"18\"\u003e\u003c\/line\u003e\u003cline x1=\"6\" y1=\"6\" x2=\"18\" y2=\"18\"\u003e\u003c\/line\u003e\u003c\/svg\u003e \u003cspan\u003ePage frozen since August 2024, with no repository update and no follow-up channel for the project\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"18\" y1=\"6\" x2=\"6\" y2=\"18\"\u003e\u003c\/line\u003e\u003cline x1=\"6\" y1=\"6\" x2=\"18\" y2=\"18\"\u003e\u003c\/line\u003e\u003c\/svg\u003e \u003cspan\u003eNo way to contact the authors from the site\u003c\/span\u003e\n\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\u003c\/section\u003e\n\u003csection class=\"block soft\" id=\"pricing\"\u003e\u003cdiv class=\"head\"\u003e\n\u003cspan class=\"eyebrow\"\u003ePricing\u003c\/span\u003e\u003ch2\u003ePricing \u0026amp; Plans\u003c\/h2\u003e\n\u003c\/div\u003e\n\u003cp class=\"rich justify\" style=\"color:var(--gv3-ink-2);margin:0 0 18px;max-width:80ch\"\u003eNo pricing is published, and none is to be expected: GameNGen is a published research result rather than a commercial offering, so there is no paid plan, no free plan and no free trial, for the simple reason that no plan of any kind exists. Consulting the project page and the gameplay videos is free of charge and requires no account, and the arXiv paper is freely accessible under a CC BY 4.0 licence. The only cost implied is indirect, and borne by anyone attempting to reproduce the work: TPU-class hardware for real-time inference, together with a complete reimplementation from the paper.\u003c\/p\u003e\n\u003cdiv class=\"pricing-extras\"\u003e\u003cdiv class=\"affiliate-note\"\u003ePrices and plans listed above may evolve. Always check the official pricing page before subscribing.\u003c\/div\u003e\u003c\/div\u003e\u003c\/section\u003e\n\u003csection class=\"block\" id=\"trust\"\u003e\u003cdiv class=\"head\"\u003e\n\u003cspan class=\"eyebrow\"\u003eTrust \u0026amp; Privacy\u003c\/span\u003e\u003ch2\u003eData, GDPR \u0026amp; hosting\u003c\/h2\u003e\n\u003cp class=\"lead\"\u003eA consolidated view of how GameNGen handles your data.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"trust-grid\"\u003e\n\u003cdiv class=\"col\"\u003e\n\u003cdiv class=\"trust-card\"\u003e\n\u003ch3\u003eGDPR overview\u003c\/h3\u003e\n\u003cp class=\"justify\" style=\"color:var(--gv3-ink-2);margin:0\"\u003eThere is \u003cb\u003eno mention of the GDPR anywhere on the site\u003c\/b\u003e. Searches for \"GDPR\", \"privacy\" and \"policy\" return no first-party occurrence. There is no privacy policy, no cookie banner, no consent notice, no Article 27 representative, no data protection officer and no contact address of any kind. That silence is not a statement of compliance: the page does load Google Analytics (gtag.js, id G-CS1E2LNG5X) and an embedded YouTube iframe, third-party trackers set without prior information or any consent mechanism. Readers in the European Union should weigh the gap between the total absence of documentation and the trackers actually loaded. GameNGen is, in context, a static GitHub Pages project page with no user account and no data collection of its own, but that context does not replace the missing notice.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"trust-card\"\u003e\n\u003ch3\u003eWho owns the data?\u003c\/h3\u003e\n\u003cp class=\"justify\" style=\"color:var(--gv3-ink-2);margin:0 0 12px\"\u003eNo terms of service, privacy policy or legal notice is published, so no document states who owns what. The project itself collects nothing directly: there is no account, no form and no upload, and a visitor only reads a page and watches videos. The site nevertheless loads third-party resources: Google Analytics (gtag.js, id G-CS1E2LNG5X), Google Fonts, ajax.googleapis.com, cdn.jsdelivr.net, Adobe DocumentCloud and an embedded YouTube player. Any browsing telemetry therefore falls under those third parties' own terms, not under a project policy. Hosting is GitHub Pages, so GitHub's conditions govern the site rather than anything written by the authors. The arXiv paper is released under a CC BY 4.0 licence.\u003c\/p\u003e\n\u003ch3\u003eReuse rights\u003c\/h3\u003e\n\u003cp class=\"justify\" style=\"color:var(--gv3-ink-2);margin:0\"\u003eThere is no user data to reuse, because the site produces no output for the visitor: everything on the page is pre-recorded material. The arXiv paper carries a CC BY 4.0 licence, so it may be read, quoted and redistributed with attribution, and the project page supplies a BibTeX entry for academic citation. The gameplay videos are recordings supplied by the authors, and the page states no reuse licence for them. The simulated content is DOOM, a 1993 commercial game by id Software, and the page says nothing about the legal status of that third-party work. Outside the paper itself, nothing should be assumed cleared for reuse.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"col\"\u003e\u003cdiv class=\"trust-card\"\u003e\n\u003ch3\u003eHosting summary\u003c\/h3\u003e\n\u003cp class=\"justify\" style=\"color:var(--gv3-ink-2);margin:0 0 14px\"\u003eNo data hosting information is published by the project. That is consistent with what the site does: there is no account, no form and no upload, so the authors collect no user data whose location would need to be declared. No jurisdiction, country or hosting region is announced anywhere.\n\nThe hosting of the site itself can be observed, and should not be confused with data hosting. The page is served by GitHub Pages; DNS resolves to 185.199.109.153, geolocated in the United States (San Francisco) on AS54113, Fastly, Inc. That address is anycast, so it does not designate a single data centre and gives no reliable indication of where the content is physically served from for a given visitor. Whatever applies to the site therefore comes from GitHub's infrastructure and conditions, not from any commitment made by the GameNGen authors. Third-party resources loaded by the page, such as Google Analytics, Google Fonts, jsDelivr, Adobe DocumentCloud and an embedded YouTube player, are hosted and governed by their own operators, and the project discloses nothing about them.\u003c\/p\u003e\n\u003c\/div\u003e\u003c\/div\u003e\n\u003c\/div\u003e\u003c\/section\u003e\n\u003csection class=\"block watch\" id=\"watch\"\u003e\u003cdiv class=\"head\"\u003e\n\u003cspan class=\"eyebrow\"\u003eWatch-outs\u003c\/span\u003e\u003ch2\u003eThings to keep in mind\u003c\/h2\u003e\n\u003cp class=\"lead\"\u003eRisks and trade-offs to weigh before adopting GameNGen.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003cul class=\"brand-list\" style=\"grid-template-columns:repeat(2,1fr);display:grid;gap:12px 24px\"\u003e\n\u003cli\u003eIntellectual property: the system simulates DOOM, a commercial work by id Software, and the page says nothing about the legal status of that simulation or of the videos\u003c\/li\u003e\n\u003cli\u003eNon-reproducibility: without code or weights, the results rest on the authors' word and on the ICLR review alone\u003c\/li\u003e\n\u003cli\u003eOver-reading the claim: \"real-time game engine\" can easily be taken for a ready-to-use product, which it is not\u003c\/li\u003e\n\u003cli\u003eScope of the human study: raters judged short clips, which is not evidence of general indistinguishability\u003c\/li\u003e\n\u003cli\u003eA neural engine has no explicit game state and no guaranteed rules, so inconsistencies are possible and cannot be fixed the way a code bug is fixed\u003c\/li\u003e\n\u003cli\u003ePrivacy: the page loads Google Analytics and an embedded YouTube iframe with no privacy policy and no consent request\u003c\/li\u003e\n\u003cli\u003eDisplacement of creative work if such models replaced explicit game design, with the mirror risk of depending on non-deterministic output; the page has also been unmaintained since August 2024, so the figures quoted age without warning\u003c\/li\u003e\n\u003c\/ul\u003e\u003c\/section\u003e\n\u003csection class=\"block\" id=\"setup\"\u003e\u003cdiv class=\"head\"\u003e\n\u003cspan class=\"eyebrow\"\u003eSetup\u003c\/span\u003e\u003ch2\u003eSetup \u0026amp; Integrations\u003c\/h2\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"setup-grid\"\u003e\u003cdiv class=\"setup-card\"\u003e\n\u003ch3 style=\"font-size:15px;margin-bottom:10px\"\u003eTechnical difficulty\u003c\/h3\u003e\n\u003cp class=\"justify\" style=\"color:var(--gv3-ink-2);margin:0 0 14px\"\u003eThere is no setup: nothing can be installed, configured or deployed. Consulting the project page, watching the videos and reading the paper demands no particular skill beyond the ability to follow an ML paper. Reproducing the system is a different matter entirely, since it requires training a reinforcement learning agent on DOOM, reimplementing the conditioning on past frames and actions, fine-tuning Stable Diffusion v1.4 and its latent decoder, then serving real-time inference on TPU hardware. That level of work belongs to advanced ML research teams with access to accelerators, not to end users.\u003c\/p\u003e\n\u003c\/div\u003e\u003c\/div\u003e\u003c\/section\u003e\n\u003csection class=\"block soft\" id=\"company\"\u003e\u003cdiv class=\"head\"\u003e\n\u003cspan class=\"eyebrow\"\u003eCompany\u003c\/span\u003e\u003ch2\u003eBehind GameNGen\u003c\/h2\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"company-card\"\u003e\n\u003cdiv class=\"kv\"\u003e\n\u003cdiv class=\"k\"\u003eCompany name\u003c\/div\u003e\n\u003cdiv class=\"v\"\u003e\u003cb\u003eGoogle Research\u003c\/b\u003e\u003c\/div\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"kv\"\u003e\n\u003cdiv class=\"k\"\u003eFounded\u003c\/div\u003e\n\u003cdiv class=\"v\"\u003e28\/08\/2024\u003c\/div\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"kv\"\u003e\n\u003cdiv class=\"k\"\u003eCountry of origin\u003c\/div\u003e\n\u003cdiv class=\"v\"\u003e🇺🇸 United States\u003c\/div\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"kv\"\u003e\n\u003cdiv class=\"k\"\u003eUBO\u003c\/div\u003e\n\u003cdiv class=\"v\"\u003eINFORMATION_NOT_FOUND\u003c\/div\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"kv\"\u003e\n\u003cdiv class=\"k\"\u003eUBO country\u003c\/div\u003e\n\u003cdiv class=\"v\"\u003eINFORMATION_NOT_FOUND\u003c\/div\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"kv\"\u003e\n\u003cdiv class=\"k\"\u003eDomain registrar country\u003c\/div\u003e\n\u003cdiv class=\"v\"\u003e🇺🇸 United States\u003c\/div\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\u003c\/section\u003e\n\u003csection class=\"block\" id=\"resources\"\u003e\u003cdiv class=\"head\"\u003e\n\u003cspan class=\"eyebrow\"\u003eOfficial links\u003c\/span\u003e\u003ch2\u003eResources\u003c\/h2\u003e\n\u003cp class=\"lead\"\u003eAll the official URLs gathered for verification and reference.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"res-grid\"\u003e\u003ca class=\"res\" href=\"https:\/\/gamengen.github.io\/\" target=\"_blank\" rel=\"noopener\"\u003e\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003ccircle cx=\"12\" cy=\"12\" r=\"10\"\u003e\u003c\/circle\u003e\u003cpath d=\"M12 2a14.5 14.5 0 0 0 0 20 14.5 14.5 0 0 0 0-20\"\u003e\u003c\/path\u003e\u003cpath d=\"M2 12h20\"\u003e\u003c\/path\u003e\u003c\/svg\u003e\u003cdiv class=\"res-text\"\u003e\n\u003cdiv\u003eHomepage\u003c\/div\u003e\n\u003cdiv class=\"url\"\u003egamengen.github.io\u003c\/div\u003e\n\u003c\/div\u003e\u003c\/a\u003e\u003c\/div\u003e\u003c\/section\u003e\n\u003csection class=\"block\" id=\"faq\"\u003e\u003cdiv class=\"head\"\u003e\n\u003cspan class=\"eyebrow\"\u003eFAQ\u003c\/span\u003e\u003ch2\u003eFrequently asked questions\u003c\/h2\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"faq\"\u003e\n\u003cdetails\u003e\u003csummary\u003eCan I play GameNGen? \u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"12\" y1=\"5\" x2=\"12\" y2=\"19\"\u003e\u003c\/line\u003e\u003cline x1=\"5\" y1=\"12\" x2=\"19\" y2=\"12\"\u003e\u003c\/line\u003e\u003c\/svg\u003e\u003c\/summary\u003e\u003cdiv class=\"ans justify\"\u003eNo. There is no playable demo and no binary. The project page only offers recorded gameplay videos.\u003c\/div\u003e\u003c\/details\u003e\u003cdetails\u003e\u003csummary\u003eAre the code or the model weights available? \u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"12\" y1=\"5\" x2=\"12\" y2=\"19\"\u003e\u003c\/line\u003e\u003cline x1=\"5\" y1=\"12\" x2=\"19\" y2=\"12\"\u003e\u003c\/line\u003e\u003c\/svg\u003e\u003c\/summary\u003e\u003cdiv class=\"ans justify\"\u003eNo. There is no release. The GameNGen GitHub organisation hosts a single repository, the website itself.\u003c\/div\u003e\u003c\/details\u003e\u003cdetails\u003e\u003csummary\u003eWhich game does it simulate, and how fast does it run? \u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"12\" y1=\"5\" x2=\"12\" y2=\"19\"\u003e\u003c\/line\u003e\u003cline x1=\"5\" y1=\"12\" x2=\"19\" y2=\"12\"\u003e\u003c\/line\u003e\u003c\/svg\u003e\u003c\/summary\u003e\u003cdiv class=\"ans justify\"\u003eIt simulates DOOM, the 1993 title, at more than 20 frames per second on a single TPU.\u003c\/div\u003e\u003c\/details\u003e\u003cdetails\u003e\u003csummary\u003eWhat image quality does it reach? \u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"12\" y1=\"5\" x2=\"12\" y2=\"19\"\u003e\u003c\/line\u003e\u003cline x1=\"5\" y1=\"12\" x2=\"19\" y2=\"12\"\u003e\u003c\/line\u003e\u003c\/svg\u003e\u003c\/summary\u003e\u003cdiv class=\"ans justify\"\u003eA PSNR of 29.4 on next-frame prediction, which the authors compare to lossy JPEG compression.\u003c\/div\u003e\u003c\/details\u003e\u003cdetails\u003e\u003csummary\u003eWhich model is it built on, and how was it trained? \u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"12\" y1=\"5\" x2=\"12\" y2=\"19\"\u003e\u003c\/line\u003e\u003cline x1=\"5\" y1=\"12\" x2=\"19\" y2=\"12\"\u003e\u003c\/line\u003e\u003c\/svg\u003e\u003c\/summary\u003e\u003cdiv class=\"ans justify\"\u003eOn Stable Diffusion v1.4, repurposed and conditioned on past frames and past actions. Training runs in two phases: a reinforcement learning agent plays DOOM and records its episodes, then a diffusion model learns to predict the next frame from those recordings.\u003c\/div\u003e\u003c\/details\u003e\u003cdetails\u003e\u003csummary\u003eHow is auto-regressive drift avoided? \u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"12\" y1=\"5\" x2=\"12\" y2=\"19\"\u003e\u003c\/line\u003e\u003cline x1=\"5\" y1=\"12\" x2=\"19\" y2=\"12\"\u003e\u003c\/line\u003e\u003c\/svg\u003e\u003c\/summary\u003e\u003cdiv class=\"ans justify\"\u003eGaussian noise is added to the encoded context frames during training, so the network learns to correct information sampled in previous frames. The latent decoder is fine-tuned on top of that.\u003c\/div\u003e\u003c\/details\u003e\u003cdetails\u003e\u003csummary\u003eWho produced this work? \u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"12\" y1=\"5\" x2=\"12\" y2=\"19\"\u003e\u003c\/line\u003e\u003cline x1=\"5\" y1=\"12\" x2=\"19\" y2=\"12\"\u003e\u003c\/line\u003e\u003c\/svg\u003e\u003c\/summary\u003e\u003cdiv class=\"ans justify\"\u003eDani Valevski and Yaniv Leviathan (Google Research), Moab Arar (Tel Aviv University) and Shlomi Fruchter (Google DeepMind), with equal contribution.\u003c\/div\u003e\u003c\/details\u003e\u003cdetails\u003e\u003csummary\u003eWhere can I read the paper and how do I cite it? \u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"12\" y1=\"5\" x2=\"12\" y2=\"19\"\u003e\u003c\/line\u003e\u003cline x1=\"5\" y1=\"12\" x2=\"19\" y2=\"12\"\u003e\u003c\/line\u003e\u003c\/svg\u003e\u003c\/summary\u003e\u003cdiv class=\"ans justify\"\u003eThe paper is arXiv 2408.14837, accepted at ICLR 2025. A BibTeX entry is provided at the bottom of the project page.\u003c\/div\u003e\u003c\/details\u003e\u003cdetails\u003e\u003csummary\u003eIs there any pricing, account or support? \u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"12\" y1=\"5\" x2=\"12\" y2=\"19\"\u003e\u003c\/line\u003e\u003cline x1=\"5\" y1=\"12\" x2=\"19\" y2=\"12\"\u003e\u003c\/line\u003e\u003c\/svg\u003e\u003c\/summary\u003e\u003cdiv class=\"ans justify\"\u003eNone at all. This is a research project page, with no commercial offering and no support channel.\u003c\/div\u003e\u003c\/details\u003e\u003cdetails\u003e\u003csummary\u003eDoes the site publish a privacy policy? \u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"14\" height=\"14\"\u003e\u003cline x1=\"12\" y1=\"5\" x2=\"12\" y2=\"19\"\u003e\u003c\/line\u003e\u003cline x1=\"5\" y1=\"12\" x2=\"19\" y2=\"12\"\u003e\u003c\/line\u003e\u003c\/svg\u003e\u003c\/summary\u003e\u003cdiv class=\"ans justify\"\u003eNo, none, even though the page loads Google Analytics and an embedded YouTube iframe.\u003c\/div\u003e\u003c\/details\u003e\n\u003c\/div\u003e\u003c\/section\u003e\n\u003csection class=\"block conclusion\" id=\"conclusion\"\u003e\u003cdiv class=\"head\"\u003e\n\u003cspan class=\"eyebrow\"\u003eConclusion\u003c\/span\u003e\u003ch2\u003eShould you pick GameNGen?\u003c\/h2\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"rich justify\" style=\"max-width:80ch\"\u003e\n\u003cp\u003eGameNGen matters as a demonstration rather than as a tool: it shows that a diffusion model, properly conditioned and stabilised, can take the place of a game engine and hold interactive real time. That result opens a credible path towards interactive world models, well beyond video games.\u003c\/p\u003e\n\u003cp\u003eCredibility is solid on the publication side. The paper was accepted at ICLR 2025, the authors work at Google Research, Google DeepMind and Tel Aviv University, and the figures given are precise enough to be argued with: more than 20 frames per second on a single TPU, a PSNR of 29.4 on next-frame prediction, and a human rater study in which participants do only slightly better than chance on short clips.\u003c\/p\u003e\n\u003cp\u003eThe structural limitation is just as clear. \u003cb\u003eNothing is delivered\u003c\/b\u003e: no code, no weights, no checkpoint, no playable demo. A reader cannot verify, extend or appropriate the work without a full reimplementation, and real-time inference assumes TPU-class hardware. The project page has not moved since August 2024, offers no contact channel and publishes no legal document, while loading Google Analytics and an embedded YouTube player.\u003c\/p\u003e\n\u003cp\u003eThe real audience follows from that: researchers, ML engineers and game developers keeping watch on the field. There is nothing here for an end user looking for something to play, nor for a team looking for something to ship. This record should be read the way a scientific publication is read, as a dated and signed landmark whose contribution is an idea and a set of measurements, not a piece of software to put to work.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"cta-row\"\u003e\u003ca class=\"btn btn-primary\" href=\"https:\/\/gamengen.github.io\/\" target=\"_blank\" rel=\"noopener\"\u003e\u003csvg viewbox=\"0 0 24 24\" fill=\"none\" stroke=\"currentColor\" stroke-width=\"2\" stroke-linecap=\"round\" stroke-linejoin=\"round\" width=\"16\" height=\"16\"\u003e\u003cpath d=\"M15 3h6v6\"\u003e\u003c\/path\u003e\u003cpath d=\"M10 14 21 3\"\u003e\u003c\/path\u003e\u003cpath d=\"M18 13v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6\"\u003e\u003c\/path\u003e\u003c\/svg\u003e Visit GameNGen\u003c\/a\u003e\u003c\/div\u003e\u003c\/section\u003e\n\u003c\/div\u003e","brand":"Google Research","offers":[{"title":"Default Title","offer_id":58253673595211,"sku":null,"price":0.0,"currency_code":"EUR","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/1038\/0905\/7099\/files\/official-gamengen-by-guidaio_61c70935-4f6a-4d6e-baf1-1d765af83d6e.png?v=1787695056","url":"https:\/\/guidaio.com\/products\/gamengen","provider":"Guidaio","version":"1.0","type":"link"}