I want to turn plain text prompts into full, explorable open-world maps. The heart of the job is fusing a large-language model with a procedural pipeline so that, for example, the phrase “snow-capped range sheltering a frontier town beside a frozen lake” instantly spawns a playable scene inside a modern game engine. Scope • Integrate an LLM (GPT-4, Llama 3 or similar) that interprets user prompts and outputs structured generation parameters • Build or adapt a terrain generator that converts those parameters into heightmaps, biomes, vegetation, roads and settlement placeholders • Stream the data into either Unity or Unreal—whichever you are fastest with—and auto-populate meshes, materials and lighting so the world loads in-engine with no manual touch-up Deliverables 1. Source code with clear comments and a brief README explaining model calls, seed control and performance considerations 2. A sample executable or project file that loads at least one generated open world from a single-sentence prompt 3. Short loom-style video (2-3 min) demonstrating prompt-to-world flow Acceptance Criteria • Prompt latency under 60 seconds for a 2 km × 2 km map on a mid-range PC • Generated terrain is traversable end-to-end without obvious gaps or floating assets • All code compiles and runs out-of-the-box following the README steps If you have existing terrain or city-building systems we can plug in, mention them—speed is a plus as long as quality holds.