Model
Trends
.ai
Model
Trends
.ai
die besten Open-Source-KI-Modelle
Autoplay
Lokal ausführen
H3_Combat_V2.safetensors
148 MB · SafeTensor · bf16
Action Director.txt
Other
Herunterladen
Director of “One Against Many”.txt
Other
Herunterladen
Example Workflow.json
0 MB · Other
Herunterladen
Downloads werden von Civitai bereitgestellt. Manche Ersteller verlangen zuerst eine Anmeldung bei Civitai.
AD
Auf Graydient installieren / ausführen
Mitgliedschaft erforderlich
<combat-h3>
Über dieses Modell
MINIMAX H3 · LoRAs
Erstellt von
FourBunny
auf Civitai
By FourBunny - MiniMax H3 LoRA for focusing not only on faster combat, but on better action continuity, hit reactions, cumulative damage, spatial movement, and physical cause-and-effect. It also works for adding richer body movement, finer sound details, and stronger shot-to-shot continuity to character performance. Author notes: For normal combat scenes, simply use prfight1 as a trigger. When the fight needs a clear finisher, knockout, takedown, or decisive ending, add prfin1 after prfight1. Add prslow1 when you want a brief slow-motion emphasis on the decisive finishing strike. (Noted to not be a guarantee to work)
Notizen des Erstellers
🚀 Run It on RunningHub 🎁 New users on RunningHub International can get 1000 RH Coins after registration. 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 Future FourBunny LoRAs will also be released and updated on RunningHub. 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 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。 Core Physics & Motion Logic / 核心物理与动作逻辑 - 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. - 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。 - Hit reactions now persist and accumulate—sustained torso bending, lowered guards, and impaired mobility directly build toward convincing heavy finishers. - 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。 Combat Skillset & Staging / 动作技法与镜头调度 - Upgrades include continuous rotational kicks, sustained grappling and takedowns, dynamic 3-stage environmental collisions, and multi-phase fall trajectories. - 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。 - Supports cinematic 1vMany choreography via staggered attacker entrances, spatial lane control, continuous protagonist-centered tracking, and motivated cross-cut motion continuity. - 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。 Dialogue Performance & Benchmarks / 文戏表现与测试数据 - 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. - 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。 Workflow & Trigger Usage / 工作流与触发词使用 - For optimal flow, structure prompt timing as ACTION → DIALOGUE; start with No Trigger for maximum anatomical accuracy, add prfight2 for higher intensity, or prfight2, prfin1 for heavy stunt impact. - 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 prfight2，重终结/特技受击则使用 prfight2, prfin1。 - Includes two specialized combat-director workflows (General & 1vMany) with Chinese instructions that support prompt output in any target language. - 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。 Direct the fight first. Write the prompt second. 先把战斗导演好，再去写提示词。 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. About This LoRA｜关于这个 LoRA 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 two people fighting、a woman fights a man 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. Trigger Words｜触发词 prfight1 — General Combat｜普通战斗 普通战斗只需要使用 prfight1。 For normal combat scenes, simply use prfight1. prfight1, prfin1 — Finisher / Knockout｜终结 / 击倒 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 prfight1 后加入 prfin1。 When the fight needs a clear finisher, knockout, takedown, or decisive ending, add prfin1 after prfight1. prfight1, prfin1, prslow1 — Slow-Motion Finisher｜慢镜终结 需要在最后一击加入短暂的慢动作强调时，再加入 prslow1。 Add prslow1 when you want a brief slow-motion emphasis on the decisive finishing strike. 注意：只加入 prslow1 并不代表模型一定会自动产生慢动作。 Important: simply adding prslow1 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. An Easy Way to Write Better H3 Fight Prompts｜最简单的高质量提示词写法 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
Triggerwörter
prfight2
prfin1
7,153 Downloads · 459 Likes auf Civitai
SHA256
5B3EDB09E9D6029439BADF9BC9DB4BA4A355AE679B0810888DE82C4793C0BAC8
ModelTrends.ai-Modell-ID
#27406
Inhalt melden
Beispielbilder
Prompts und Einstellungen, geteilt von den Leuten, die diese Bilder auf Civitai erstellt haben. Wähle ein Bild, um zu sehen, wie es entstanden ist.
Einstellungen
Schritte
12
Guidance
1
Größe
1440x832
Prompt
For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced. integrated_multimodal_description: [Shot 1] The scene opens from <Picture 1>, preserving the three young women standing close together outdoors on a city street in winter, with European-style buildings and bare trees in the background under soft daylight. The camera has a natural handheld smartphone selfie shakiness, as if held by one of the girls at arm's length. The girl on the left with long straight black hair, a thick gray scarf, and a dark coat (S1) shivers slightly, pulling her scarf tighter around her neck with one hand. She looks directly into the camera lens with a bright, slightly complaining expression and says: <d>[Chinese] 今天有点冷啊，快看镜头啊！我们三个谁最好看啊？</d> The girl in the middle with chestnut-brown hair, a gray coat with bow buttons, and a white phone with a bear charm (S2) turns her gaze toward the camera, a playful smile forming on her lips. The girl on the right with a single braid, a gray sweater, and a gold necklace (S3) continues looking down at her dark phone, smiling faintly. [Shot 2] At 00:05.000, the handheld camera wobbles slightly as the girls adjust their positions. The middle girl (S2) raises her hand in a cute peace-sign gesture beside her face, tilts her head playfully, and laughs with a cheeky, confident expression, saying: <d>[Chinese] 当然是我啦！</d> She winks one eye playfully after the line. The left girl (S1) reacts with a mock-offended expression, opening her mouth in playful protest and lightly nudging the middle girl's shoulder with her elbow. The right girl (S3) finally looks up from her phone toward the camera, an amused smile spreading across her face as she shakes her head at the other two. [Shot 3] At 00:10.000, the right girl (S3) laughs warmly, her eyes crinkling with genuine amusement, and says in a light, scolding-but-affectionate tone: <d>[Chinese] 两个臭屁精！别拍了，赶快吃火锅去啦！</d> She puts her dark phone away into her coat pocket and starts to turn her body toward the street, gesturing with her hand for the others to follow. The middle girl (S2) giggles and lowers her peace-sign hand, while the left girl (S1) laughs and nods in agreement. The handheld camera begins to lower and tilt as if the person filming is putting the phone down, the three girls starting to move together toward the street, their shoulders brushing as they walk off, still laughing. The frame ends with a slight downward tilt as the selfie video concludes. overall_soundscape: Soft outdoor city street ambience continues throughout — distant traffic hum, occasional car passing, faint wind rustling through the bare tree branches, and the low murmur of pedestrians in the background. The girls' voices are clear and close, captured by the phone's microphone with a slight natural reverb from the open street. Subtle fabric rustling is audible as they adjust their coats and scarves, and the soft click of a phone lock button is heard near the end as the right girl puts her phone away. Light, cheerful laughter from all three girls overlaps naturally throughout the conversation. non_diegetic_music: N/A
Einstellungen
Größe
960x544
Prompt
V2BASE
Einstellungen
Größe
1376x768
Prompt
integrated_multimodal_description: [Shot 1] A surreal martial-arts action scene. A woman with exactly three legs stands in the center of a large training courtyard. She is wearing a traditional white karate gi with a black belt. She is surrounded by a group of masked ninjas dressed entirely in black, they are bare handed. The woman suddenly raises her body onto her middle leg, using it as a single central pivot. Her two other legs extend horizontally in opposite directions, forming a three-point spinning shape. She begins spinning rapidly around her middle leg like a human whirlwind. The middle leg remains planted firmly on the ground and acts as the central axis, while the two outer legs stay fully extended and rotate around her. Several ninjas attempt to rush toward her from different directions, but each time one approaches, one of her two extended legs swings around and delivers a powerful spinning kick, sending the ninja tumbling backward. Other ninjas try to approach from different angles, only to be knocked away by the rotating legs. The choreography is clear and continuous: one leg remains the stationary pivot while the other two legs remain extended and sweep around her in a wide circular motion. The woman maintains perfect balance and moves with exaggerated martial-arts precision. The scene is spectacular and absurd, played completely straight, with dynamic camera movement emphasizing the circular motion and the increasingly desperate ninjas surrounding her. overall_soundscape: Ambient noise. non_diegetic_music: N/A
Einstellungen
Größe
720x960
Prompt
V2BASE
Einstellungen
Größe
416x736
Prompt
the girl from example.png finally meets ryu from streetfighter in a dark ally. the girl proves far more superior in hand to hand combat since she is fighting in her own comfyui environment using a modified workflow.
Einstellungen
Schritte
25
Seed
265159317415717
Größe
1472x1472
Prompt
prfight2, high-end claymation combat sequence, tactile stop-motion texture. A square golden-brown waffle fighter with short articulated limbs and glossy chocolate-syrup eyes is surrounded in a narrow wooden breakfast arena by a surging horde of round pancakes; only three foreground pancake warriors have crude short stubby articulated limbs and receive individual hit reactions while the background mass closes as one blurred threat. The waffle is on the defensive throughout: every foreground exchange begins with a pancake attack, followed by a visible block, parry, or dodge from the waffle, then one heavier counter that opens space to his right. Sequential single-target exchanges only; the camera pans with him to the right as he moves through the gaps created by his counters. A first pancake lunges from ahead-right with an overhead batter-punch at his head; the waffle ducks under it and turns its momentum aside with a low forearm parry, then drives one heavy upward elbow strike into the pancake’s exposed center; it bends violently upward, cracks along its edge, syrup spurts outward, and collapses flat behind him, opening a narrow gap to his right. A second pancake immediately surges through that gap with a horizontal batter-punch at his chest; the waffle absorbs it against his raised guard in a loud wet snap, recoils half a step right but stays upright, then answers instantly with a low forearm strike under the attacker’s rim; the pancake flips over, batter splashes in an exaggerated arc, lands flat and crumples with visible clay deformation, clearing another lane segment to his right. A third pancake lunges overhead from the closing horde with a broadside edge punch; the waffle ducks under it and counters before it can recover by driving one heavy straight punch into its center; the attacker cracks diagonally, tumbling pieces flatten behind him as he continues rightward. The remaining blurred pancake mass recoils and collapses in broad waves behind him, ending with a low-angle victory pose at the right edge of the cleared arena as flattened pancakes trail behind him. Dynamic tracking shot panning sharply right with his defensive movement throughout, impact-synced camera kicks, exaggerated squash-and-stretch while preserving coherent silhouettes, heavy physical consequences after every attack and counter. Audio: Heavy punch and guard-clash impacts from both sides: wet slaps, batter snaps, floppy pancake squelches on incoming pancake attacks; deep cartoonish bass thuds and crunchy waffle cracks on the waffle’s defensive counters; thick syrup splatter whooshes; low-end impact drops synced to every landed counter hit. Percussive rhythm, no melody.
KI-Modelle für Bild-, Video- und Texterzeugung
combat h3 ist ein LoRA für die Familie MiniMax H3 und wird auf ModelTrends.ai gelistet, einem reinen Lesekatalog für Open-Source-KI-Modelle für Bild, Video und Text. Vergleiche es mit anderen MiniMax H3-Modellen, prüfe den Hitzewert, um den Trend zu sehen, und öffne es auf PirateDiffusion oder BitVector, um es auszuprobieren.
Nach Modellfamilie stöbern
Alle Modelle ansehen