<h1 id="lingbot-world-v2-14b-causal-fast-gguf-(8gb-vram-ready)">LingBot-World-v2 14B causal-fast — GGUF (8GB-VRAM ready)</h1><h3 id="work-in-progress!">WORK IN PROGRESS!</h3>GGUF quantizations of <a target="_blank" rel="ugc" href="https://huggingface.co/robbyant/lingbot-world-v2-14b-causal-fast">robbyant/lingbot-world-v2-14b-causal-fast</a> for <a target="_blank" rel="ugc" href="https://github.com/RealRebelAI/ComfyUI_Rebels_LingBotWorld">ComfyUI_Rebels_LingBotWorld</a> — an action-controlled world model running on consumer GPUs (tested: RTX 3070 8GB / 16GB RAM).You author a camera/movement track (or type one into the Action Builder node); the model renders the video that track produces from your start image, with genuine action following. Offline generation, chunked causal sampling, 4 distilled steps per chunk.<h2 id="files">Files</h2>FileNotes<code>LingBot-World-14B-Q4_K_S.gguf</code> (~11.7GB)tested tier; Q5-bumped v-projections keep effective bpw ~6.7other tiers (Q4_K_M…Q8_0)quality ladder; RAM-streamed, VRAM use is unchangedAlso required: <code>Wan2.1_VAE.pth</code>, a UMT5-XXL GGUF encoder (loads via ComfyUI-GGUF CLIPLoader, type <code>wan</code>)<h2 id="settings-that-matter-(8gb)">Settings that matter (8GB)</h2><ul><li>Resolution preset 256×448 (default) or 320×544; world-memory window 6+2</li><li>The KV cache is the world memory: it scales with window × resolution (upstream 18+6 @ 480×832 ≈ 21GB — does not fit consumer cards; the sampler pre-checks and refuses instead of hanging)</li><li>frame_num 4n+1; start at 21; export 16 fps</li></ul><h2 id="license">License</h2>CC BY-NC-SA 4.0 (inherited from upstream): non-commercial, attribution, share-alike. Quantization is a format conversion only. Quants + nodes by <a target="_blank" rel="ugc" href="https://huggingface.co/realrebelai">RealRebelAI</a>.

Files with SHA256: d414e312625bb5cfb95a1d0c8942352790639b8e4a0fd32c8b658097a01120ee