🚀 Run It on RunningHub🎁 New users on RunningHub International can get 1000 RH Coins after registration.<h3 id="new-workflow-name:-bunny_selflift_12gb-direct-1080p_high-motion-+-lora-sharp_hybrid-workflowbunny_selflift_12gb1080p_+lora_"><a target="_blank" rel="ugc" href="https://www.runninghub.ai/zh-cn/post/2102271264323829761/?inviteCode=lk4qa7rh">NEW Workflow name: BUNNY_SelfLift_12GB Direct 1080P_High Motion + LoRA Sharp_Hybrid Workflow 工作流名称：BUNNY_SelfLift_12GB直出1080P_高动态+LORA不模糊_混合工作流</a></h3>Tested hardware: RTX 4070 Ti 12GB VRAM. 实测硬件：RTX 4070 Ti 12GB 显存。A 10-second H3 video at around 1080P / 2.0MP takes approximately 15 minutes locally on 12GB VRAM. 10秒 H3 视频在约1080P / 2.0MP下，本地12GB显存实测生成时间约 15分钟。The same 10-second test takes approximately 7 minutes on RunningHub. 同样的10秒测试，在 RunningHub 上大约 7分钟完成<a target="_blank" rel="ugc" href="https://www.runninghub.ai/zh-cn/post/2095927034992533505/?inviteCode=lk4qa7rh">🎬 BUNNY H3 High-Dynamic Two-Pass Latent Upscale Workflow</a><a target="_blank" rel="ugc" href="https://www.runninghub.ai/zh-cn/post/2093321887345819650/?inviteCode=lk4qa7rh ">▶ Start / End Frame One-Click Workflow</a><a target="_blank" rel="ugc" href="https://www.runninghub.ai/zh-cn/post/2093328847935950849/?inviteCode=lk4qa7rh ">▶ Multi-Reference One-Click Workflow</a><a target="_blank" rel="ugc" href="https://www.runninghub.ai/zh-cn/user-center/2092843565587107841/model?inviteCode=lk4qa7rh">📦Bunny Official RunningHub Model Page</a>Future FourBunny LoRAs will also be released and updated on RunningHub.<hr /><h3 id="highly-recommended-for-complex-action:-pair-this-lora-with-bunny-h3-conditioning-bridge-for-better-multi-character-logic-action-ownership-and-scene-stability.-bunny-h3-conditioning-bridge"><a target="_blank" rel="ugc" href="https://civarchive.com/models/2936635/bunny-h3-conditioning-bridge-or-multi-character-logic-and-scene-stability">Highly recommended for complex action: pair this LoRA with BUNNY H3 Conditioning Bridge for better multi-character logic, action ownership, and scene stability. 复杂动作场景强烈推荐搭配 BUNNY H3 Conditioning Bridge，可进一步改善多人互动逻辑、动作归属和场景稳定性。</a></h3><hr /><h2 id="combat-base-v2-is-a-major-upgrade-over-v1-focusing-not-only-on-faster-combat-but-on-better-action-continuity-hit-reactions-cumulative-damage-spatial-movement-and-physical-cause-and-effect.">Combat Base V2 is a major upgrade over V1, focusing not only on faster combat, but on better action continuity, hit reactions, cumulative damage, spatial movement, and physical cause-and-effect.</h2><h3 id="combat-base-v2-is-no-longer-just-a-combat-lora-it-now-works-for-both-action-and-dialogue-scenes-adding-richer-body-movement-finer-sound-details-and-stronger-shot-to-shot-continuity-to-character-performance.combat-base-v2-lora">Combat Base V2 is no longer just a combat LoRA — it now works for both action and dialogue scenes, adding richer body movement, finer sound details, and stronger shot-to-shot continuity to character performance. Combat Base V2 不再只是一个武打 LoRA——它现在可文可武，在文戏中也能增加更丰富的身体动作、更细致的声音表现，以及更强的镜头与状态连续性。</h3><hr /><h3 id="minimax-h312gb-vramuniversal-fl2varef2va-base-model-+-low-sigma-combat-second-pass-workflow"><a target="_blank" rel="ugc" href="https://civarchive.com/models/2869434/minimax-h312gb-vramuniversal-fl2varef2va-base-model-low-sigma-combat-second-pass-workflow?modelVersionId=3241900">【MiniMax H3】12GB VRAM｜Universal FL2VA/REF2VA Base Model + Low-Sigma Combat Second-Pass Workflow</a></h3><h2 id="core-physics-and-motion-logic">Core Physics &amp; Motion Logic / 核心物理与动作逻辑</h2><ul><li>Combat Base V2 moves away from "attack → reset" loops into a continuous physical chain where missed strikes carry momentum, blocks change direction smoothly, and limbs separate cleanly without sticking.</li><li>Combat Base V2 摆脱了“攻击 → 复位”的回合制循环，形成持续的运动链，招式落空可延续动量，格挡可自然顺势变向，且接触点脱离干净不黏连。</li><li>Hit reactions now persist and accumulate—sustained torso bending, lowered guards, and impaired mobility directly build toward convincing heavy finishers.</li><li>受击状态具备持续性与累计效应，躯干弯曲、护架变慢及移动受限等伤害会自然保留，让最后的重击终结更加合理成立。</li></ul><h2 id="combat-skillset-and-staging">Combat Skillset &amp; Staging / 动作技法与镜头调度</h2><ul><li>Upgrades include continuous rotational kicks, sustained grappling and takedowns, dynamic 3-stage environmental collisions, and multi-phase fall trajectories.</li><li>全面升级连续旋转腿法、持续近身缠抱摔投、三阶段环境碰撞（人体→环境→材质），以及拥有完整轨迹的特技跌落表现。</li><li>Supports cinematic 1vMany choreography via staggered attacker entrances, spatial lane control, continuous protagonist-centered tracking, and motivated cross-cut motion continuity.</li><li>支持电影级“一对多”围攻调度，利用交替切入、环境通道压迫、主角中心跟随镜头以及跨镜动作连贯性提升画面表现。</li></ul><h2 id="dialogue-performance-and-benchmarks">Dialogue Performance &amp; Benchmarks / 文戏表现与测试数据</h2><ul><li>Internal same-seed dialogue tests revealed +19.5% in-shot motion, +34.2% audio RMS, and +22.4% edge sharpness, significantly enhancing physical acting, breathing, and vocal dynamics outside of fight scenes.</li><li>同 Seed 文戏测试显示镜内动作量增加 +19.5%、Audio RMS 提升 +34.2%、边缘锐度提升 +22.4%，大幅增强了武戏之外的肢体表演、呼吸与声音动态。</li></ul><h2 id="workflow-and-trigger-usage">Workflow &amp; Trigger Usage / 工作流与触发词使用</h2><ul><li>For optimal flow, structure prompt timing as ACTION → DIALOGUE; start with No Trigger for maximum anatomical accuracy, add <code>prfight2</code> for higher intensity, or <code>prfight2, prfin1</code> for heavy stunt impact.</li><li>建议将提示词结构设为“动作 → 对白”；优先使用 No Trigger 换取最稳动作与人体，需要强攻击性时加入 <code>prfight2</code>，重终结/特技受击则使用 <code>prfight2, prfin1</code>。</li><li>Includes two specialized combat-director workflows (General &amp; 1vMany) with Chinese instructions that support prompt output in any target language.</li><li>附带“通用”与“一对多”两个专用武戏导演工作流，内置中文说明并支持生成任意语言的最终 Prompt。</li></ul>Direct the fight first. Write the prompt second. 先把战斗导演好，再去写提示词。<hr />Upgrade Notes：测试发现：同一镜头包含打斗和台词时，必须先完整完成击打动作再说台词，否则容易退回底模的慢动作表现。Test finding: When combat and dialogue are in the same shot, complete the attack before the dialogue, or the motion may revert to the base model’s slower behavior.<hr /><h2 id="about-this-lora-lora">About This LoRA｜关于这个 LoRA</h2>先说明一下：这不是一个魔法 LoRA，也不会自动帮你编排完整的武打动作。 First of all: this is not a magic LoRA, and it will not automatically choreograph a complete fight for you.如果你只写 <code>two people fighting</code>、<code>a woman fights a man</code> 之类非常简单的 Prompt，模型很可能只是让两个人高速乱挥拳，最后变成一段很快但没有逻辑的“王八拳”。 You still need to tell the model who attacks first, what they do, where the strike lands, how the opponent dodges or reacts, and why the next attack can naturally follow.Prompt 负责决定“他们怎么打”，LoRA 负责增强“这些动作怎么表现”。 The prompt decides what the fighters do, while the LoRA enhances how those actions are performed.这个 Combat LoRA 前后训练和测试了好几个版本，也尝试过不同的数据比例、战斗速度、受击反应以及终结动作训练方向。 This Combat LoRA went through several training and testing versions with different dataset balances, combat-speed behavior, hit reactions, and finisher-focused training.有些旧版本确实能让战斗明显变快，但高权重下也容易破坏 H3 本身不错的动作流畅度，让视频出现倍速感、模糊或者没有恢复过程的连续乱打。 Some earlier versions could make fights noticeably faster, but at higher weights they could also damage H3's native fluidity and produce sped-up motion, blur, or endless attacks without proper recovery.目前发布的这一版，是我实际测试下来相对最平衡、也比较好用的一版。 The version released here is currently one of the most balanced and practical versions from my testing.它不再单纯追求“打得更快”，而是重点强化 连续战斗、命中反馈、身体失衡、后仰、踉跄、击倒、终结逻辑以及攻击产生的实际结果。 Instead of simply making fights faster, it focuses more on continuous combat, impact reactions, loss of balance, recoil, staggering, knockdowns, finishing logic, and clearer physical consequences after attacks.我的目标是尽量保留 H3 本身已经不错的动作连续性，同时让每一次有效攻击看起来更有重量、更有反馈，也更有“打中了以后真的发生了什么”的感觉。 The goal is to preserve H3's existing motion continuity while making successful attacks feel heavier, more reactive, and more consequential.<hr /><h2 id="trigger-words">Trigger Words｜触发词</h2><h3 id="prfight1-general-combat"><code>prfight1</code> — General Combat｜普通战斗</h3>普通战斗只需要使用 <code>prfight1</code>。 For normal combat scenes, simply use <code>prfight1</code>.<h3 id="prfight1-prfin1-finisher-knockout"><code>prfight1, prfin1</code> — Finisher / Knockout｜终结 / 击倒</h3>当这一段战斗需要一个明确的终结、KO、击倒或决定性收尾时，在 <code>prfight1</code> 后加入 <code>prfin1</code>。 When the fight needs a clear finisher, knockout, takedown, or decisive ending, add <code>prfin1</code> after <code>prfight1</code>.<h3 id="prfight1-prfin1-prslow1-slow-motion-finisher"><code>prfight1, prfin1, prslow1</code> — Slow-Motion Finisher｜慢镜终结</h3>需要在最后一击加入短暂的慢动作强调时，再加入 <code>prslow1</code>。 Add <code>prslow1</code> when you want a brief slow-motion emphasis on the decisive finishing strike.注意：只加入 <code>prslow1</code> 并不代表模型一定会自动产生慢动作。 Important: simply adding <code>prslow1</code> does not guarantee that slow motion will automatically appear.你还需要在 Prompt 里明确告诉模型：最后一击在命中瞬间进入短暂慢动作，然后恢复正常速度。 You also need to explicitly tell the model that the decisive final strike receives a brief slow-motion emphasis at the moment of impact, then normal speed resumes.<hr /><h2 id="an-easy-way-to-write-better-h3-fight-prompts">An Easy Way to Write Better H3 Fight Prompts｜最简单的高质量提示词写法</h2>如果你不想自己手动写复杂的动作编排，一个非常简单的方法是直接使用 MiniMax H3 官方的两个 Prompt Writing Guide 文件。 If you do not want to manually write complex choreography, an easy method is to use the two official MiniMax H3 Prompt Writing Guide files.把这两个官方提示词模板文件一起上传给 Gemini。 Upload both official prompt-writing guide files to Gemini.然后告诉 Gemini：按照这两个官方文件的规则，用六段式结构为 H3 编写视频提示词，并告诉它视频长度、剧情和你想要的战斗内容。 Then tell Gemini: follow the rules from these two official files and write an H3 video prompt using a six-part structure, while specifying the video duration, story, and desired fight choreography.通常你只需要告诉 Gemini：几秒、什么剧情、谁打谁、最后想怎么结束，它就可以按照官方模板帮你扩展成一段完整得多的 H3 武打提示词。 Usually, you only need to tell Gemini the duration, basic story, who fights whom, and how you want the sequence to end, and it can expand that into a much more complete H3 fight prompt using the official structure.这也是我目前比较推荐的使用方式：不要让模型自己猜怎么打，而是先让 Gemini 帮你完成动作编排，再把结果交给 H3 + Combat LoRA。 This is also my recommended workflow: do not make the video model guess the choreography; let Gemini build the action sequence first, then send the resulting prompt to H3 + Combat LoRA.<hr /><h2 id="compatibility">Compatibility｜兼容性说明</h2>目前我只在 MiniMax H3 FL2VA 上完整测试过这个 LoRA。 At the moment, I have only fully tested this LoRA with MiniMax H3 FL2VA.其他 H3 模型、不同生成模式或其他工作流我暂时没有进行系统性测试。 I have not systematically tested it with other H3 variants, generation modes, or workflows.所以在非 FL2VA 环境中能否获得相同的效果，我目前无法保证。 Therefore, I cannot currently guarantee identical behavior outside the FL2VA setup.

Files with SHA256: 8cec1cf4cbc2cb75e9a62ad702bc56dfee26c1efef8641d0418fe9821335eaef