🎁 Register on RunningHub &amp; Get 1,000 RH Coins:🎁 注册 RunningHub 领 1,000 RH 币：BUNNY H3 High-Dynamic FL2VA / REF2VA One-Click Workflow :<a target="_blank" rel="ugc" href="https://www.runninghub.ai/zh-cn/post/2095927034992533505/?inviteCode=lk4qa7rh">https://www.runninghub.ai/zh-cn/post/2095927034992533505/?inviteCode=lk4qa7rh</a>RH Generation Speed / 运行效率：On RunningHub, generating a 1080P (0.9MP × 1.5) 10-second video takes ~12 minutes.<h2 id="overview">📌 Overview / 工作流简介</h2>This workflow is specially optimized for MiniMax H3 high-dynamic video generation. It solves common issues such as noise, blur, and detail instability when applying secondary latent upscaling alongside <a rel="ugc" href="https://civarchive.com/models/2853878/minimax-h3-combat-base-fight-motion-impact-drama-booster">BUNNY COMBAT LoRAs</a>.本工作流专为 MiniMax H3 高动态视频生成 进行深度优化，精准解决了配合 BUNNY 动作 LoRA 进行二次放大时容易出现的噪点、画面模糊和细节不稳定等痛点。<ul><li>Output Capability / 画面表现：Supports direct output up to 1080P / 最高支持 1080P 直出。</li><li>12GB VRAM Optimization / 12GB 本地显存优化：<ul><li>Recommended Stage 1 base resolution: 0.3MP with 1.5× Latent Upscaling.</li><li>Note: Max achievable duration depends on VRAM, aspect ratio, and the number of reference images.</li><li>本地 12GB 显存推荐设置：一采使用 0.3MP 基础分辨率 + 1.5 倍 Latent 放大。</li><li>提示：实际可生成时长会根据显存大小、画幅比例及参考图数量有所不同。</li></ul></li></ul><hr /><h2 id="how-to-use">🚀 How to Use / 使用指南</h2><ol><li>Prompt &amp; Duration / 提示词与时长：Enter your prompt and set the target video duration. / 首先填写提示词和生成时长。</li><li>Resolution &amp; Upscale Scale / 分辨率与放大倍率：Set the Stage 1 base resolution and latent upscale scale (recommended starting value: 1.5×). / 设置一采基础分辨率与 Latent 放大倍率（推荐从 1.5 倍开始）。</li><li>Reference Images / 参考图设置：Load images directly when using FL2VA or REF2VA. For Text-to-Video (T2V), bypass or disconnect all reference image nodes. / 使用 FL2VA 或 REF2VA 时直接加载参考图；使用文生视频（T2V）时，请旁路（Bypass）或断开所有参考图节点。</li><li>Conditioning / 条件节点：Keep both Conditioning nodes set to AUTO. / 两路 Conditioning 节点保持 AUTO 默认设置即可。</li><li>LoRAs Fine-tuning / LoRA 调整：Replace or adjust the Stage 1 and Stage 2 LoRAs according to your subject. / 一采与二采的 LoRA 可根据生成题材自行更换或微调。</li><li>Stage 2 Enforcement / 二采必须开启：Stage 2 reconstructs the upscaled latent—do NOT disable it. / 二采负责重构放大后的 Latent 潜空间，切勿关闭。</li></ol><hr /><h2 id="required-custom-nodes-and-models">🧩 Required Custom Nodes &amp; Models / 必需依赖项</h2><h3 id="1.-custom-nodes">1. Custom Nodes / 必需第三方节点</h3><ul><li>MiniMax H3 Audio T8: <a target="_blank" rel="ugc" href="https://github.com/T8mars/comfyui-minimax-h3-audio-T8">GitHub Link</a></li><li>ComfyUI-KJNodes: <a target="_blank" rel="ugc" href="https://github.com/kijai/ComfyUI-KJNodes">GitHub Link</a></li><li>ComfyUI-Easy-Use (Optional: Used for post-generation VRAM cleanup only / 仅用于可选的生成后显存清理功能): <a target="_blank" rel="ugc" href="https://github.com/yolain/ComfyUI-Easy-Use">GitHub Link</a></li></ul><h3 id="2.-required-latent-upscaler-latent">2. Required Latent Upscaler / 必需 Latent 放大模型</h3><blockquote>⚠️ Must use the exact model linked below / 必须使用指定版本：</blockquote><ul><li>Model Download / 模型下载：<a target="_blank" rel="ugc" href="https://huggingface.co/LBH-123-AI/Minimax_h3_latent_Upscaler/resolve/main/minimax_h3_latent_upscaler_3d_fp16.safetensors">minimax_h3_latent_upscaler_3d_fp16.safetensors</a></li><li>Target Path / 存放路径：<code>ComfyUI/models/latent_upscale_models/</code></li><li>SHA-256（节点实际强制 / runtime enforced）: <code>043E5A48E161610EF6C3EA974645220354D06FA618ABCA15F76D084812EB55C2</code>加载器还会检查 322 个 tensor 的键、形状和类型，防止旧版、重打包、转换版或同名异构 checkpoint 生成坏 latent。/ The loader also validates 322 tensor keys, shapes and dtypes to reject old, repacked, converted, or structurally different same-name checkpoints before they can produce corrupted latents.</li></ul><h2 id="credits-and-open-source-spirit">❤️ Credits &amp; Open Source Spirit / 致谢与开源共建</h2><ul><li>Special Thanks to T8 for the underlying MiniMax H3 repair logic and two-pass sampling setup.特别感谢 T8 老师 提供 MiniMax H3 底层修复与双采样逻辑支持。</li><li>Thanks to LBH-123-AI for training and sharing the MiniMax H3 3D latent upscaler model.感谢 LBH-123-AI 提供 MiniMax H3 3D 潜空间放大模型。</li></ul>✨ This is how an open-source community moves forward: creators build great workflows using my LoRAs, and I optimize those workflows based on my deep understanding of BUNNY LoRAs. Everyone is welcome to continue building upon this version. Every contribution helps our technology and community grow!✨ 开源社区的魅力就在于不断的互助与进化：其他创作者基于我的 LoRA 制作出优秀的工作流，我再结合自己对 BUNNY LoRA 特性的理解对其进行针对性优化。非常欢迎大家在这个版本的基础上继续改进与创作，每一次微小的优化，都在推动技术与社区共同进步！⚠️ Disclaimer / 注意事项This download includes the ComfyUI workflow JSON file only. Third-party custom node source code and model weights are not included and must be installed separately.本资源仅包含 ComfyUI 工作流文件，不包含第三方节点源码及模型权重，请按上述说明自行安装。P.S. Hey, if I slap a VSR super-resolution node on the end to bump this to 2K, I should probably list this workflow for 800K Buzz too, right? Just checking the 2026 market rate. 😉P.S. 顺便问一句：如果我再顺手挂个 VSR 超分节点给它升到 2K 分辨率，我是不是也该把这个工作流标价 800K Buzz 挂出来卖了？纯粹打听一下 2026 年的市场行情。😉

Files with SHA256: de7dda3916cd75a6dbed4c4e35b4b78bb5111ce1e3d7d5cc793411b854a0c883