WanGP v18.81 lets you build a world from text or an image and walk through it with the keyboard
WanGP v18.81 adds a Worlds gallery where you create a world from a text description with Hy-Nomas or from an image with EVOKE, then explore it with the keyboard, save your progress and record the walk as a video. The developer says 8 GB of VRAM is enough to start, thanks to the new MMGP v4 memory manager.
What it does
Hy-Nomas 8B starts from text and remembers earlier views to keep places and characters consistent; you can walk, turn the camera or type an event such as "the fox sits down", and choose how much history it remembers. EVOKE Turbo 14B starts from an image and keeps a memory of the scene's depth so places keep their layout when you return, at 640×384. You can also capture a view to use in another generation.
Worth knowing
The release notes warn not to expect miracles: near-real-time exploration depends on the model and a fast, recent GPU, and EVOKE moves take longer to generate. The same release adds the H3 PTD RefAsymOPD 8-step LoRA (about 11 GB of extra RAM), new H3 accelerator LoRAs, two MOSS-Image editors and faster quantized KV caches, and the built-in Deepy Prime assistant can now search the web (off by default), install LoRAs from Civitai and Hugging Face, and run video restore, outpaint and upscale batches.