<h3 id="(new)-one-click-triton-and-sageattention-installation-guide:-https:civitai.comarticles12248-im1f8yo7p">(NEW) One-Click Triton &amp; SageAttention installation guide: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/12248">https://civarchive.com/articles/12248</a></h3><h1 id="v2.2-updated-(pending)-w06f0fv55">V2.2 UPDATED (PENDING)</h1>This update was tested with 8GB VRAM, removes the confusion between i2v and t2v, adds convenient resolution settings (thanks @redhome) and several other handy features. It also fixes a critical bug in shift order and adds perfectly support for WAN Lora.<h2 id="recommended-version-q3uye6548">Recommended Version</h2><ul><li>ComfyUI v0.3.19 or higherWan does not work on versions below that.</li><li>ComfyUI-Manager v3.26 or higherThere is an issue with missing nodes and conflict in previous versions</li></ul><h2 id="require-nodes-(works-on-install-missing-nodes)-fl3of4prx">Require Nodes (Works on Install Missing Nodes)</h2><ul><li>ComfyUI-KJNodes (need to update. Switch Version -&gt; latest(nightly), doesn't work 1.0.6 version or lower)</li><li>ComfyUI-GGUF (If you have other gguf custom nodes, remove them)</li><li>ComfyUI-GIMM-VFI (need to installed cuda toolkit 12.x)</li><li>ComfyUI-Easy-Use</li><li>rgthree-comfy</li><li>ComfyUI-Florence2</li><li>ComfyUI-Impact-Pack</li><li>ComfyUI-Custom-Scripts</li><li>ComfyUI-VideoHelperSuite</li><li>ComfyUI-Frame-Interpolation</li><li>ComfyUI-HunyuanVideoMultiLora</li></ul><h3 id="v2-update-notes-shskekqqo">v2 Update Notes</h3><ul><li>Applied TeaCache</li><li>Added Florence2 auto-captioning</li><li>Added GIMM VFI</li><li>Added LoRA Loader</li><li>Introduced shift parameter</li><li>Optimized workflow</li></ul><h1 id="require-models-628pqxlt6">Require Models</h1><h3 id="t2v-quant-model-ap4rhsxlr">T2V Quant Model</h3>🔗 <a target="_blank" rel="ugc" href="https://huggingface.co/city96/Wan2.1-T2V-14B-gguf/blob/main/wan2.1-t2v-14b-Q4_K_M.gguf">Download</a> 📂 Save to: <code>models/diffusion_models</code>or<h3 id="t2v-fastest-model-(works-low-vram)-8xjd2izbs">T2V Fastest Model (Works low vram)</h3>🔗 <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_t2v_1.3B_bf16.safetensors">Download</a> 📂 Save to: <code>models/diffusion_models</code><h3 id="i2v-quant-model-4cds2oqwd">I2V Quant Model</h3>🔗 <a target="_blank" rel="ugc" href="https://huggingface.co/city96/Wan2.1-I2V-14B-480P-gguf/blob/main/wan2.1-i2v-14b-480p-Q4_K_M.gguf">Download</a> 📂 Save to: <code>models/diffusion_models</code><h3 id="clip-do-not-use-another-clip-18frzm0sb">CLIP DO NOT USE ANOTHER CLIP</h3>🔗 <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors">Download</a> 📂 Save to: <code>models/clip</code><h3 id="clip-vision-qlyqm9bmc">CLIP-VISION</h3>🔗 <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/clip_vision/clip_vision_h.safetensors">Download</a> 📂 Save to: <code>models/clip_vision</code><h3 id="vae-mbcd3qhfe">VAE</h3>🔗 <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors">Download</a> 📂 Save to: <code>models/vae</code><h3 id="upscale-model-0ijoxowu9">UPSCALE MODEL</h3>🛠 Recommended Model: <code>4x_foolhardy_Remacri</code> download via ComfyUI Model Manager

Wan Video 2.1 Native Workflow

Fixed an issue where sampling would proceed even though nothing had been modified.Optimized workflowAdded LoRA pipeline node

Files with SHA256: c9dbcdf3ed31dbd093a6aa58e6a1b334f831536b0a5f0cb14dc1005e268957a9