{"rewrite":{"id":"r_dd7e2215344e879e2f215e89","clusterId":"c_101032523de10c36f4241b5c","slug":"square-enix-details-atogis-3d-generation-ai-at-cedec2026","model":"deepseek-v4-flash","headline":"Square Enix Details ATOGiS 3D Generation AI at CEDEC2026","summary":"Square Enix presented ATOGiS, its proprietary 3D generation AI, at CEDEC2026. The system combines an image-to-3D geometry AI with a retopology AI, both trained on roughly 60,000 in-house weapon and background assets to avoid licensing problems. AI \u0026 Engine Development Division member Taketo Araki led the session, covering dataset selection, a staged retopology learning method, and three pre-tests run before large-scale training on eight H100 GPUs.","whyItMatters":"Square Enix built a 3D generation AI trained only on its own assets, pairing a TRELLIS.2-based geometry model with a proprietary staged retopology architecture, to put its rarely reused asset library to work.","webCardHtml":"\u003cp\u003eSquare Enix\u0026#39;s 3D assets are designed to fit each work\u0026#39;s world view, so they rarely get reused. To make use of them, the company built its own generation AI. The ATOGiS dataset came to about 60,000 in-house weapon and background assets, with complex structures, textures, and externally licensed items excluded. Polygons over 10,000 were cut to limit GPU load.\u003c/p\u003e\u003cp\u003eThe retopology AI learns in stages: vertices first, then edges, then faces. Vertices are converted to voxel grids, edges are guided by a matrix, and faces are filled by a standard algorithm. The geometry AI customizes the TRELLIS.2 architecture, converting input images to normal maps and adding a geometry normal loss to speed learning. Before large-scale training, the team ran three tests: a learning expression test, a single overfitting test, and a category learning test.\u003c/p\u003e","blueskyPost":"Square Enix presented ATOGiS, its in-house 3D generation AI, at CEDEC2026. Trained on about 60,000 proprietary weapon and background assets, it pairs an image-to-3D geometry model with a staged retopology AI. Final training ran on eight H100 GPUs at around 300M parameters.","twitterPost":"Square Enix presented ATOGiS, its in-house 3D generation AI, at CEDEC2026. Trained on about 60,000 proprietary weapon and background assets, it pairs an image-to-3D geometry model with a staged retopology AI. Final training ran on eight H100 GPUs at around 300M parameters.","threadsPost":null,"newsletterBlurb":"Square Enix detailed ATOGiS, its in-house 3D generation AI, at CEDEC2026. The system pairs an image-to-3D geometry model with a retopology AI, both trained on roughly 60,000 proprietary weapon and background assets. The session covered dataset selection, staged retopology learning, and the pre-tests run before large-scale training on eight H100 GPUs.","attributionJson":"[{\"source\":\"GameBusiness.jp\",\"url\":\"https://www.gamebusiness.jp/article/2026/09/01/27850.html\",\"title\":\"Square Enix Develops 3D Generation AI Trained on In-House Assets [CEDEC2026]\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":6313,"outputTokens":4431,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1788350641,"createdAt":"2026-09-02T12:01:59.000Z","publishedAt":"2026-09-02T12:02:25.000Z","updatedAt":"2026-09-02T12:02:25.000Z"},"cluster":{"id":"c_101032523de10c36f4241b5c","canonicalTitle":"自社3Dアセットで学習した3D生成AIを研究開発する…スクウェア・エニックスの取り組みとは？【CEDEC2026】","representativeArticleId":"a_a55bad5db8223edb83f3a42b","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[\"Square Enix\"],\"people\":[],\"type\":\"news\",\"domain\":\"industry\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-09-01T08:00:04.000Z","lastSeenAt":"2026-09-01T08:00:04.000Z","updatedAt":"2026-09-02T12:02:24.000Z"},"attribution":[{"source":"GameBusiness.jp","url":"https://www.gamebusiness.jp/article/2026/09/01/27850.html","title":"自社3Dアセットで学習した3D生成AIを研究開発する…スクウェア・エニックスの取り組みとは？【CEDEC2026】"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":["Square Enix"],"people":[],"type":"news","domain":"industry","is_roundup":false},"keyFacts":null}