Perplexity 团队用 Fable 5.2 开发浏览器版月球火星探索游戏 Getsumen Express
Perplexity 内部团队用 Fable 5.2 做了款能在浏览器直接玩的月球火星探索游戏,建模靠 Tripo 和 Meshy,地形音效全是代码生成,总花费约 730 美元。想看 AI 编程工具能干到什么程度,点开链接就能玩。
Perplexity 联合创始人 Arav Srinivas 展示了团队成员用 Fable 5.2 完整开发的 3D 探索游戏 Getsumen Express,玩家可在浏览器中通过 spacesim.pplx.app 游玩。游戏支持在月球和火星两个场景中步行和驾驶探测器,可扫描阿波罗着陆点、Curiosity 与 Viking 1 等真实地标,也能在火星上用等离子卡宾枪射击外星人,并通过星际特快列车在两颗星球间往返。
技术上,游戏基于 Three.js 运行 WebGPU,并带 WebGL2 回退方案。车站和列车用 Blender 配 Python 脚本搭建,宇航员、探测器、卡宾枪和外星人则通过 Tripo 和 Meshy 的 image-to-3D 生成后再在 Blender 中修整。地形、岩石、基地、地标和全部音效均由代码程序化生成。
整个项目在租用的云 GPU 上用 Chrome 逐个版本测试,57 秒的 1080p60 预告片也在此完成录制。总成本约 73,000 credits(约 730 美元),按 Fable 5.2 xhigh 档位计费。
Teammate on Fable 5.2 built Getsumen Express, a Moon and Mars exploration game you can play in your browser at spacesim.pplx.app . You walk and drive a rover across both worlds, scan real landmarks like the Apollo landing sites, Curiosity and Viking 1, shoot aliens on Mars with a plasma carbine, and take an intergalactic express train between the two. It runs on Three.js with WebGPU and a WebGL2 fallback. Teammate built the stations and the train in Blender with Python scripts, and made the astronaut, rover, carbine and alien with image-to-3D (Tripo and Meshy), then cleaned them up in Blender. The terrain, rocks, home base, landmarks and all the sound are procedural, generated by code. Teammate tested every build in Chrome on a rented cloud GPU and recorded this 57-second trailer there at 1080p60. Total cost: about 73,000 credits (~$730) on Fable 5.2 xhigh.
This report is categorized under AITOP's "Products" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. 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 Fable 5.2, Perplexity, Tripo, Meshy.
Source attribution: curated from Aravind Srinivas (@AravSrinivas). Cross-source corroboration: 4 independent sources have covered this development (including 歸藏(guizang.ai), kimmonismus, Andrew Curran), confirming industry-wide signal strength.