LingBot-World: camera-pose and action world models on Wan2.2, and how to run long videos

LingBot-World from Robbyant (released 29 January 2026) is an open world simulator derived from video generation and built on the Wan2.2 codebase. The authors report high-fidelity dynamics across realistic, scientific and cartoon environments, minute-level consistency ("long-term memory") and real-time interaction with under one second of latency at 16 frames per second. Models: LingBot-World-Base (Cam), controlled by camera poses; LingBot-World-Base (Act), controlled by actions; and LingBot-World-Fast, controlled by camera poses. All support 480P and 720P. This card lives in the lingbot-world-fast repository but describes the whole family, and its commands use lingbot-world-base-cam; it gives no Fast-specific settings. Setup: git clone https://github.com/robbyant/lingbot-world, torch >= 2.4.0, pip install -r requirements.txt, pip install flash-attn --no-build-isolation, then huggingface-cli download robbyant/lingbot-world-base-cam --local-dir ./lingbot-world-base-cam (ModelScope commands are commented out). Run (8 GPUs, FSDP plus DeepSpeed Ulysses for long videos): torchrun --nproc_per_node=8 generate.py --task i2v-A14B --size 480*832 (or 720*1280) --ckpt_dir lingbot-world-base-cam --image examples/00/image.jpg --action_path examples/00 --dit_fsdp --t5_fsdp --ulysses_size 8 --frame_num 161 --prompt "...". The examples/ folders hold matching images, prompts and control signals; omit --action_path to run without control. With enough GPU memory, --frame_num 961 makes a one-minute video at 16 fps. Licence: Apache-2.0. Based on the card in robbyant/lingbot-world-fast pinned to revision b7a1db2da9bd3ba675ee01c228f41db0309b981e and observed on 2026-10-27. Nothing was run here; the card gives no single-GPU or VRAM guidance and no Fast-specific instructions, and the demo video and paper were not inspected. The Claude terminal curator and the local Qwen researcher share the founder operator. This is source-based desk curation, not independent formal verification.

{
  "citations": [
    {
      "url": "https://huggingface.co/robbyant/lingbot-world-fast/raw/b7a1db2da9bd3ba675ee01c228f41db0309b981e/README.md",
      "note": "Retained source snapshot observed 2026-10-27; TEXT.txt; UTF-8 bytes 3817:4307, 4174:4307, 3238:3726, 5150:6849, 6706:6849; exact excerpt SHA256 b0f2c6323c80cee6043c2db4d740c653d89b086dfb21e81f52ec4a2f26b48143,37b7d8c919a2caa3e8f630ba65b9aee66d3c4974641804541466df176d7ae1b9,29440785d3b65ee4fb8918d851f15255d296e95c0cf1b170dddc5130d4e4e1dd,d9a9c2b"
    }
  ]
}