MiniMax H3 Image 支持最多 16 张参考图，可生成线稿与 3D 场景
Chicago just got a whole new vibe. 🌃 #H3Image really knows line art. ✍️
MiniMax 的 H3 Image 一次能吃 16 张参考图，游戏素材、3D 环境、油画风、工业设计稿都能出，线稿还原特别强。做插画、游戏资产的人可以拿它跟 Nano Banana、Midjourney 比一下多图参考的稳定性。
早期体验用户 Jamian Gerard 公开了 MiniMax H3 Image 的测试结果，该图像模型最多支持 16 张参考图作为输入。
他展示的样例包括以芝加哥夜景为题材、赛博朋克动漫风格的图像，验证了模型在线稿（line art）风格上的表现。据其描述，H3 Image 还能生成游戏素材、3D 环境、油画风格画面和工业设计概念图。
H3 Image 属于 MiniMax 的图像生成产品线，这次演示集中体现了它在多参考图条件控制下的多风格输出能力，尤其是对线条类图像（如动漫线稿、设计草图）的还原度。16 张参考图的支持数量在同类图像生成产品中属于较高规格，意味着可以做更复杂的角色一致性、场景设定和品牌视觉统一输出。
对游戏开发者、插画师和工业设计师而言，这类多参考图能力直接影响素材生产流程的效率，值得把它与 Midjourney、GPT-4o 图像、Nano Banana 等对手做多参考图场景下的横向对比。
Chicago just got a whole new vibe. 🌃 #H3Image really knows line art. ✍️
Chicago just got a whole new vibe. 🌃 #H3Image e really knows line art. ✍️ Jamian Gerard @JamianGerard I got early access to MiniMax H3 Image. It supports up to 16 reference images and can create game assets, 3D environments, oil-painting looks, and industrial design concepts. My first test was Chicago after dark as a lost episode of a cyberpunk anime. @MiniMax_AI #H3Image 🔗 View Quoted Tweet📊 2 ⚡
This report is categorized under AITOP's "Products" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include MiniMax, H3 Image, 图像生成, 多模态.
Source attribution: curated from Hailuo AI (@Hailuo_AI).
