<h3 id="loop-video">這是一個用於生成Loop Video的工作流。因為我使用首尾幀時發現，當首幀與尾幀的圖片都是一樣時，影片動態感時常過於保守，所以我改變了生成流程。</h3><h3 id="abbcacb-greaterc-greatera">這個工作流的原理是，先用上傳的圖片生成一段影片，將其平分成兩段，分成A影片與B影片，並用B影片的尾幀當作C影片的首幀，用A影片的首幀當作C影片的尾幀，並將影片以B-&gt;C-&gt;A的方式拼接，這樣就避免使用同張圖當作首尾幀而出現動態太過保守的問題。</h3><h3 id="i2v">不過實際上，這沒能突破I2V的限制，我只能使用圖片當作首尾幀，最後拚接有可能出現動態不同的情況，這我沒辦法很好解決。</h3><h3>同時工作流也支持使用已經生成過的影片，將其分兩段並生成循環過度幀。</h3><hr /><p></p><h3 id="this-is-a-workflow-for-generating-loop-videos.-when-using-the-first-and-last-frames-i-found-that-if-both-frames-were-identical-the-video's-dynamic-feel-often-became-overly-conservative.-therefore-i-modified-the-generation-process.">This is a workflow for generating loop videos. When using the first and last frames, I found that if both frames were identical, the video's dynamic feel often became overly conservative. Therefore, I modified the generation process.</h3><h3 id="the-principle-of-this-workflow-is-to-first-generate-a-video-from-the-uploaded-image-divide-it-into-two-equal-segmentsvideo-a-and-video-bthen-use-the-final-frame-of-video-b-as-the-opening-frame-of-video-c-and-the-opening-frame-of-video-a-as-the-closing-frame-of-video-c.-the-videos-are-then-stitched-together-in-the-order-b-greaterc-greatera.-this-approach-avoids-the-overly-conservative-motion-that-occurs-when-using-the-same-image-for-both-the-first-and-last-frames.">The principle of this workflow is to first generate a video from the uploaded image, divide it into two equal segments—Video A and Video B—then use the final frame of Video B as the opening frame of Video C and the opening frame of Video A as the closing frame of Video C. The videos are then stitched together in the order B-&gt;C-&gt;A. This approach avoids the overly conservative motion that occurs when using the same image for both the first and last frames.</h3><h3 id="however-in-practice-this-doesn't-overcome-the-limitations-of-i2v.-i'm-still-restricted-to-using-images-for-the-first-and-last-frames-which-means-the-final-stitched-video-may-exhibit-inconsistent-motion.-this-is-an-issue-i-haven't-been-able-to-resolve-effectively.">However, in practice, this doesn't overcome the limitations of I2V. I'm still restricted to using images for the first and last frames, which means the final stitched video may exhibit inconsistent motion. This is an issue I haven't been able to resolve effectively.</h3><h3 id="the-workflow-also-supports-using-pre-generated-videos-splitting-them-into-two-segments-and-generating-looping-transition-frames.">The workflow also supports using pre-generated videos, splitting them into two segments and generating looping transition frames.</h3><p></p>

Wan 2.2 I2V 14B Loop video

<p>Looking forward to receiving questions and making revisions.</p><p>The model you need:</p><p>ComfyUI/</p><p>└── models/</p><p>    ├── diffusion_models/</p><p>    │   ├── DasiwaWAN22I2V14BTastysinV8_q8High.gguf</p><p>    │   └── DasiwaWAN22I2V14BTastysinV8_q8Low.gguf</p><p>    │</p><p>    ├── vae/</p><p>    │   └── wan_2.1_vae.safetensors</p><p>    │</p><p>    ├── text_encoders/</p><p>    │   └── umt5_xxl_fp8_e4m3fn_scaled.safetensors</p><p>    │</p><p>    └── upscale_models/</p><p>        └── RealESRGAN_x2.pth</p><p><a target="_blank" rel="ugc" href="https://civitai.com/models/2269796--">https://civitai.com/models/2269796--</a>&gt; Wan 2.2 I2V GGUF</p><p><a target="_blank" rel="ugc" href="https://huggingface.co/Wan-AI/Wan2.1-T2V-14B/tree/main">https://huggingface.co/Wan-AI/Wan2.1-T2V-14B/tree/main</a> --&gt;VAE2.1</p><p><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">https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors</a> --&gt; CLIP</p><p><a target="_blank" rel="ugc" href="https://huggingface.co/ai-forever/Real-ESRGAN/tree/main">https://huggingface.co/ai-forever/Real-ESRGAN/tree/main</a> --&gt; Upscale Model</p><p>The custom node you need:</p><p>ComfyUI-KJNodes</p><p>ComfyUI-ReservedVRAM</p><p>Comfyui-Memory_Cleanup</p><p>WhiteRabbit</p><p>rgthree-comfy</p><p>ComfyUI-Easy-Use</p><p>ComfyUI-GGUF</p><p>ComfyUI-VideoHelperSuite</p>

Files with SHA256: fcbefa93e437339477d862dfcfbcd5182af6ec235a372f88cc464a9e5294a1c5