wan2.2-Remix

<!-- Telegram Link --> <p align="center" style="margin-top: 20px; margin-bottom: 20px;"> <a href="https://t.me/wan22remix" target="_blank" rel="noopener"> <img src="https://img.shields.io/badge/Telegram-Group-blue?logo=telegram" alt="Join Telegram Group" style="width: 320px; height: 55px;"/> </a> <br/> <span style="font-size: 16px; font-weight: bold; display: inline-block; margin-top: 8px;"> 👉 Join our Telegram group for updates and feedback </span> </p> <section> <h2>Model Overview — Wan2.2-Remix</h2> <p> Wan2.2-Remix is a <strong>research and creative model</strong> designed for producing <strong>short, imaginative video clips</strong> from textual prompts. It explores human figure dynamics, body movement expression, and scene consistency, delivering smoother motion and enhanced realism in character interactions. No additional LoRA setup is required. </p> <ul> <li><strong>High-noise model:</strong> <a href="https://huggingface.co/lightx2v/Wan2.2-Lightning/tree/main/Wan2.2-T2V-A14B-4steps-250928-dyno" target="_blank" rel="noopener"> Lightx2v Wan2.2 Lightning dyno version </a> (quantized to <strong>fp8</strong>) </li> <li><strong>Low-noise model:</strong> <a href="https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models" target="_blank" rel="noopener"> Comfy-Org Wan 2.2 ComfyUI Repackaged </a> (fp16, quantized to <strong>fp8</strong>) </li> <li><strong>Enhancements:</strong> Blended with open-source LoRA resources and customized training data to enrich motion, body posture, and conceptual fidelity. </li> <li><strong>Integration:</strong> Already bundled with <code>lightx2v</code>, no extra setup required</li> <li><strong>Recommended CLIP:</strong> <a href="https://huggingface.co/NSFW-API/NSFW-Wan-UMT5-XXL/tree/main" target="_blank" rel="noopener"> NSFW-Wan-UMT5-XXL </a> </li> </ul> <p> This release is currently <strong>beta</strong>. While functional, it is not yet perfect—feedback and testing are highly appreciated. </p> </section> <hr/> <section> <h2>Changelog</h2> <h4>v0.9</h4> <ul> <li>High-noise model updated to <strong>Lightx2v dyno version</strong>, quantized to <strong>fp8</strong></li> <li>Low-noise model updated to <strong>Comfy-Org fp16 version</strong>, quantized to <strong>fp8</strong></li> <li>Improved <strong>character motion coherence</strong></li> <li>Enhanced <strong>natural human limb articulation</strong></li> </ul> <h4>v0.8</h4> <ul> <li>Refined body motion balance for smoother expression</li> <li>Improved anatomical rendering for both male and female figures</li> <li>Adjusted walking animation weights for realism</li> <li>Enhanced visual clarity in intimate interaction scenarios</li> </ul> <h4>v0.7</h4> <ul> <li>Introduced basic body dynamics simulation</li> <li>Added anatomical detail to human figures</li> <li>Improved facial rendering and expression consistency</li> <li>Added dynamic actions: walking, dancing, running</li> <li>Enhanced pose generation and overall gesture variety</li> </ul> </section> <hr/> <section> <h2>Disclaimer</h2> <p> This model is released for <strong>research and creative purposes only</strong>. It must not be used for illegal, unethical, or harmful activities. The author assumes no responsibility for any misuse or consequences arising from its use. </p> </section> <hr/> <section> <h2>模型概述 — Wan2.2-Remix</h2> <p> Wan2.2-Remix 是一款<strong>研究與創意模型</strong>，可根據文字提示生成 <strong>短片段的影像作品</strong>。 本模型著重於人體姿態、肢體動作表現力以及場景一致性， 讓人物動態更自然流暢，互動更加真實。 無需額外導入 LoRA。 </p> <ul> <li><strong>高躁模型：</strong> <a href="https://huggingface.co/lightx2v/Wan2.2-Lightning/tree/main/Wan2.2-T2V-A14B-4steps-250928-dyno" target="_blank" rel="noopener"> Lightx2v Wan2.2 Lightning dyno 版本 </a>（量化至 <strong>fp8</strong>） </li> <li><strong>低躁模型：</strong> <a href="https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models" target="_blank" rel="noopener"> Comfy-Org Wan 2.2 ComfyUI Repackaged </a>（fp16，量化至 <strong>fp8</strong>） </li> <li><strong>增強：</strong> 融合開源 LoRA 資源與自訂訓練數據， 提升動作表現、姿態細節與概念一致性。 </li> <li><strong>整合：</strong> 已內建 <code>lightx2v</code>，無需額外設置</li> <li><strong>推薦 CLIP：</strong> <a href="https://huggingface.co/NSFW-API/NSFW-Wan-UMT5-XXL/tree/main" target="_blank" rel="noopener"> NSFW-Wan-UMT5-XXL </a> </li> </ul> <p> 本版本目前為 <strong>測試階段</strong>，仍存在改進空間，歡迎回饋與測試。 </p> </section> <hr/> <section> <h2>更新日誌</h2> <h4>v0.9</h4> <ul> <li>高躁模型更新為 <strong>Lightx2v dyno 版本</strong>，量化為 <strong>fp8</strong></li> <li>低躁模型更新為 <strong>Comfy-Org fp16 版本</strong>，量化為 <strong>fp8</strong></li> <li>改善<strong>人物動作的協調性</strong></li> <li>增強<strong>人體肢體的自然表現</strong></li> </ul> <h4>v0.8</h4> <ul> <li>優化肢體動態平衡，使表現更自然</li> <li>改善男女角色的解剖細節呈現</li> <li>調整行走動作權重以提升真實感</li> <li>增強互動場景中的細節與清晰度</li> </ul> <h4>v0.7</h4> <ul> <li>引入基礎的肢體動態模擬</li> <li>增加人體解剖細節</li> <li>改善臉部渲染與表情一致性</li> <li>新增動作：行走、跳舞、奔跑</li> <li>提升姿態生成與動作多樣性</li> </ul> </section> <hr/> <section> <h2>免責聲明</h2> <p> 本模型僅供<strong>研究與創意用途</strong>。 禁止用於任何非法、不道德或有害的行為。 作者不承擔任何因濫用而產生的後果。 </p> </section>