This GGUF file is a direct conversion of Wan-AI/Wan2.2-T2V-A14B Since this is a quantized model, all original licensing terms and usage restrictions remain in effect. Usage The model can be used with the ComfyUI custom node ComfyUI-GGUF by city96 Place model files in ComfyUI/models/unet see the GitHub readme for further installation instructions.
Open-weight model · Text to video
minimax-h3-spatial-physics-lora
by Hpmg Jojocodex/minimax-h3-spatial-physics-lora
空间思维与物理逻辑 LoRA，基于 MiniMax-H3（Comfy-Org/MiniMax-H3）训练，让模型学会纯物体的空间关系与物理运动（碰撞、堆叠、掉落、遮挡等）。 目前还是训练和测试阶段，一些素材片段以及打标问题导致LORA还不是特别稳定。下一阶段准备修复后重新训练。目前LORA也是可以使用，强度建议0.3 ~ 0.5。纯属是在原模型基础上稍微增强一点物理反馈。 最新是重新训练到了wushuspatialphysicsclean3000pruned.safetensors 版本。…
Parameters—
Context—
Weights310.2 MB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads25.4k
Model Card
By Hpmg, published under apache-2.0, revision 476b24df28b9.
空间思维与物理逻辑 LoRA，基于 MiniMax-H3（Comfy-Org/MiniMax-H3）训练，让模型学会纯物体的空间关系与物理运动（碰撞、堆叠、掉落、遮挡等）。 目前还是训练和测试阶段，一些素材片段以及打标问题导致LORA还不是特别稳定。下一阶段准备修复后重新训练。目前LORA也是可以使用，强度建议0.3 ~ 0.5。纯属是在原模型基础上稍微增强一点物理反馈。 最新是重新训练到了wushuspatialphysicsclean3000pruned.safetensors 版本。 实测用了这个LORA，视频整体提升真实感物理的逻辑，比如物体碰撞的真实反馈。也可以用于一些打斗场景，人物真实碰撞的效果。 用了空间物理LORA的武打片段（强度 0.3） 没有用空间物理LORA的武打片段（同样提示词） 1. 下载.safetensors，放入 ComfyUI models/loras/ 2. LoraLoader 加载，strength 建议 0.8~1.0 3. 用空间/物理语言 prompt 描述物体运动 - several colored objects 多个彩色物体（CLEVRER 风格） - rigid objects 刚体 / elastic objects 弹性物体 - metal objects 金属物体 - a ball / balls 球 - billiard balls 台球 - objects 通用物体 - on a table 桌面上（PhyCo 台球场景） - in a synthetic scene 合成场景（CLEVRER 风格）…
Read Hpmg's full model card
空间物理 LoRA · Spatial & Physics LoRA (MiniMax-H3)
空间思维与物理逻辑 LoRA，基于 MiniMax-H3（Comfy-Org/MiniMax-H3）训练，让模型学会纯物体的空间关系与物理运动（碰撞、堆叠、掉落、遮挡等）。
本 LoRA 专注物体物理。
目前还是训练和测试阶段，一些素材片段以及打标问题导致LORA还不是特别稳定。下一阶段准备修复后重新训练。目前LORA也是可以使用，强度建议0.3 ~ 0.5。纯属是在原模型基础上稍微增强一点物理反馈。 最新是重新训练到了wushu_spatial_physics_clean_3000_pruned.safetensors 版本。
视频对比
实测用了这个LORA，视频整体提升真实感物理的逻辑，比如物体碰撞的真实反馈。也可以用于一些打斗场景，人物真实碰撞的效果。
用了空间物理LORA的武打片段（强度 0.3）
没有用空间物理LORA的武打片段（同样提示词）
其他测试视频 （wushu_spatial_physics_v2_3000.safetensors）
使用方法（ComfyUI）
- 下载
*.safetensors，放入 ComfyUImodels/loras/ - LoraLoader 加载，strength 建议 0.8~1.0
- 用空间/物理语言 prompt 描述物体运动
关键词库（从训练数据提炼）
① 主体/场景
- several colored objects 多个彩色物体（CLEVRER 风格）
- rigid objects 刚体 / elastic objects 弹性物体
- metal objects 金属物体
- a ball / balls 球
- billiard balls 台球
- objects 通用物体
- on a table 桌面上（PhyCo 台球场景）
- in a synthetic scene 合成场景（CLEVRER 风格）
② 物理行为
- collision / colliding 碰撞
- a single collision between objects 单次碰撞
- bouncing / rebound 弹跳/回弹
- with visible impact and bounce 可见撞击与反弹
- with minimal bounce 小幅度回弹
- rolling into a wall 滚向墙
- scattering 散射 / 四散
- momentum transfer 动量传递
- in motion 运动中的
③ 空间一致性短语（关键！）
- clear spatial layout 清晰的空间布局
- visible depth 可见的纵深
- consistent trajectories 一致的轨迹
- gravity-consistent / gravity-consistent movement 重力一致的运动
- gravity-consistent rebound trajectories 重力一致的回弹轨迹
- objects maintaining spatial relationships 物体保持空间关系
- realistic restitution 真实的恢复系数（回弹性）
- rigid collision with momentum transfer 带动量传递的刚性碰撞
负面提示词建议
训练时 neg 为空（模型未学负向），建议只加画质/物理负向，不要加与物理矛盾的内容：
blurry, distorted, low quality, jittery motion, objects passing through each other, objects floating without gravity, morphing shapes, inconsistent lighting, flickering
提示词示例（可直接使用）
碰撞类
1. several colored objects, a single collision between objects, metal objects with rigid, elastic collisions, clear spatial layout, visible depth, objects maintaining spatial relationships in a synthetic scene
2. several colored objects, objects colliding with visible impact and bounce, clear spatial layout, gravity-consistent trajectories
3. objects colliding with visible impact and bounce, clear spatial layout, gravity-consistent trajectories
弹跳/回弹类
4. Elastic objects bouncing with realistic restitution, clear spatial layout, gravity-consistent rebound trajectories
5. objects bouncing with realistic restitution, clear spatial layout, gravity-consistent rebound
6. a ball rolling into a wall and rebounding with minimal bounce, clear spatial layout, visible depth, rigid collision with momentum transfer
多物体散射/台球类
7. billiard balls colliding and scattering on a table, momentum transfer between balls, clear spatial layout, visible depth
8. several balls on a table, a ball striking the group, momentum transfer, balls scattering with realistic collision, clear spatial layout, gravity-consistent
轨迹/无碰撞类
9. Rigid objects in motion with clear spatial layout, consistent trajectories, gravity-consistent movement and visible depth
10. several colored objects, objects moving with consistent trajectories, no collision, clear spatial layout, visible depth, objects maintaining spatial relationships in a synthetic scene
自由发挥模板（替换行为词）
11. a ball thrown across the scene, [parabolic arc / gravity-consistent trajectory], clear spatial layout, visible depth
12. two balls colliding in mid-air, clear spatial layout, consistent trajectories, gravity-consistent rebound, visible depth
```
加速适配
与 MiniMax-H3 Turbo 加速 LoRA 兼容，可叠加使用。
训练信息
- 框架：ai-toolkit · Rank 16 · 分辨率 512 · 90 帧 @ 24fps
Identity and Version
- Repository
- Jojocodex/minimax-h3-spatial-physics-lora
- Publisher
- Hpmg
- Task
- Text to video
- Modality
- Video
- Library
- minimax-h3
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- 476b24df28b9b7cc5481b750469698f5cc4b0558
- First published
- 2026-08-16
- Last updated
- 2026-08-19
Files and Weights
13 files, 326.2 MB in total. The weights are 2 files totalling 310.2 MB in safetensors.
Weights2 files · 310.2 MB
Documentation1 file · 6.2 KB
Other9 files · 16.0 MB
Repository1 file · 2.2 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| wushu_spatial_physics_clean_3000_pruned.safetensors | Weights | 155.1 MB | 7d14f3701560 |
| wushu_spatial_physics_v2_1000_pruned.safetensors | Weights | 155.1 MB | c38b8f9c4361 |
| README.md | Documentation | 6.2 KB | — |
| video/MiniMax_H3_00098_.mp4 | Other | 836.1 KB | 3700db3b43d5 |
| video/MiniMax_H3_00136_.mp4 | Other | 612.8 KB | 6017e9dcf1c7 |
| video/MiniMax_H3_00138_.mp4 | Other | 1.5 MB | ef32c9fa6068 |
| video/MiniMax_H3_00140_.mp4 | Other | 1.3 MB | 4c3cdf1a83c0 |
| video/MiniMax_H3_00144_.mp4 | Other | 1.5 MB | b227d6ce1169 |
| video/MiniMax_H3_00145_.mp4 | Other | 533.5 KB | 4e27f2094922 |
| video/MiniMax_H3_00146_.mp4 | Other | 677.7 KB | 88e0eeb40cd7 |
| video/with lora.mp4 | Other | 4.4 MB | 111b798bfb05 |
| video/without lora.mp4 | Other | 4.7 MB | 570dd176f6f9 |
| .gitattributes | Repository | 2.2 KB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 310.2 MB
Download from Hpmg
Released by Hpmg through its official repository on Hugging Face. Read the license.
Built From
- Adapter of Comfy-Org/MiniMax-H3
- Derived from Comfy-Org/MiniMax-H3
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 310.2 MB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About minimax-h3-spatial-physics-lora
Can I use minimax-h3-spatial-physics-lora commercially?
Yes. minimax-h3-spatial-physics-lora is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.
Similar Models
A LoRA for MiniMax-H3 that renders joint video + synchronized stereo audio in as few as 4 sampling steps instead of the usual ~20 — a ~5× sampling speedup — and keeps getting better as you add steps. For most work, use minimaxh3turbov4step600ema.safetensors. It's the markedly better micro-detail (faces, fingers, fine texture), and the over-sharpening / plastic look of the earlier v1 (~850) line is fully resolved. v4 introduced a static-frame enhancement — a big win for static and small-motion content. The one trade-off shows up only at 4 steps with large, fast motion, where v4 can produce motion-smear / trailing ghosting (we're actively fixing this). Two things address it: - Use 6–8 steps.…
This repository contains MiniMax-H3 Turbo LoRAs converted and optimized for ComfyUI: These LoRAs accelerate MiniMax-H3 video and synchronized-audio generation by reducing the required number of sampling steps. Newly added LoRA, located in the experimental/ folder: Manual recommended sigmas: 3-step 1.0, 0.961165, 0.853333, 0.0 4-step 1.0, 0.970874, 0.907249, 0.640000, 0.0 Three LoRAs extracted from VDN-H3 8 step: The main 8-step LoRA works on both FL2VA and Ref2VA. If you are running a pruned base, choose the pruned version that corresponds to your base — the pruned versions need their own matching pruned base. Three dynamically resized BF16 LoRAs are now included. Their source weights were…
Sulphur 2 An uncensored video generation model based on LTX 2.3 supporting both t2v and i2v natively, as well as all of the other ltx 2.3 formats. Follow us on X Join our Discord Support the next version of the project, even just a few dollars would go a long way: Kofi To get started with the model, I recommend downloading either of the dev versions, (fp8mixed or bf16) and downloading the distill lora provided. By the way, I'm aware the workflows contain sulphurfinal right now, just use the lora or use the full models, don't use both at the same time. This model contains a prompt enhancer. The easiest way to get started with the prompt enhancer is by using it on lmstudio. The way to…
Wan2.1 is an open-source suite of video foundation models, compatible with consumer-grade GPUs, that excels in various video generation tasks like text-to-video, image-to-video, and video editing, even supporting visual text generation. Download models using huggingface-cli: You can also download directly from this page. This model is a derivative work of the original model licensed under the Apache 2.0 License, and is therefore distributed under the terms of the same license. Thanks to Patrick Gillespie for creating the ASCII text art tool used in this project https://patorjk.com/software/taag/ Wan-AI for the Wan model https://huggingface.co/Wan-AI/Wan2.1-VACE-1.3B…
Wan2.1 is an open-source suite of video foundation models, compatible with consumer-grade GPUs, that excels in various video generation tasks like text-to-video, image-to-video, and video editing, even supporting visual text generation. Download models using huggingface-cli: You can also download directly from this page. This model is a derivative work of the original model licensed under the Apache 2.0 License, and is therefore distributed under the terms of the same license. Thanks to Patrick Gillespie for creating the ASCII text art tool used in this project https://patorjk.com/software/taag/ Wan-AI for the Wan model https://huggingface.co/Wan-AI/Wan2.1-T2V-1.3B…