Minimax-h3_Singularity

Minimax-h3_Singularity

Generates video from text, images, references or other videos in ComfyUI with a fine-tuned fusion of MiniMax-H3, now in a V2 INT8 release. For video creators who want stronger action scenes and style control.

Made by WarmBloodAban on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: Minimax-h3_Singularity is a fine-tuned fusion of MiniMax-H3 for video generation. It supports text-to-video, image-to-video, reference-to-video and video-to-video workflows in ComfyUI. The latest V2 INT8 release adds wider action coverage, closer prompt following, a fix for audio reference conditioning and stronger visual styles.

Who it suits: It suits video creators and ComfyUI users who want clean, sharp, dynamic clips with good motion and camera control.

Industry | Who uses it | How they use it | Example
Film and video | Independent creators | Turn an image into a moving shot | A fantasy duel with particle effects
Games | Concept and trailer artists | Prototype action and spell effects | A martial arts sequence for a trailer
Marketing | Content teams | Make styled clips from a prompt | A cinematic CG product teaser

People who work outside ComfyUI can skip it, since the card describes its workflows only inside ComfyUI.

Where to use it:

• ComfyUI, for all four video workflows
• Windows, macOS or Linux, where you can download the files with Python
• An online demo on RunningHub, which the card links

Use case | Who | What they get back
Image to video | Creators | A clip that animates a still image
Action scenes | Game artists | Smoother motion in sword fights, martial arts and melee combat

How to set it up: You need these first:

• Python and the huggingface_hub package
• ComfyUI, since the card describes its workflows there
• The MiniMax-H3 base model the card says this one builds on

1. Install the Hugging Face hub package.
2. Download the model files to your computer.
3. Load the model in a ComfyUI video workflow.

pip install -U huggingface_hub
hf download WarmBloodAban/Minimax-h3_Singularity

How to use it: In your prompt, name the subject, what it does, the setting, how the camera moves and the look you want. Keep instructions consistent and put the details that matter most first. Add style keywords that match the look you want.

Visual style | Suggested prompt keywords
Photorealistic | photorealistic, natural lighting, realistic textures
3D CG | 3D CG, detailed materials, stylized 3D rendering
Cinematic CG | cinematic CG, cinematic lighting, film-quality rendering

Example: You give a still image and a prompt that names the subject, a sword fight, a mountain courtyard, a tracking shot and the cinematic CG style. The card says V2 follows such instructions more closely, so the clip should show that action with that camera move.

For newcomers and experts: A video model turns your words or a picture into a short moving clip. A first small thing to try is an image-to-video run with a short prompt that names the subject, action, setting, camera and style.

The card says the model comes from a fusion of key checkpoints, followed by long fine-tuning and three days of pruning and weight tuning to remove training artifacts. It was tuned on HDR video data to cut blur in fast action. V2 asks for prompts within about 5,000 characters, and short, well-structured ones are welcome.

Write a prompt with subject, action and style
Pick a ComfyUI video workflow
Load the V2 INT8 model
Run and review the clip

Good to know:

• Versions: V2 is released in INT8 format, and Hugging Face lists 5 quantized versions made by other people
• Languages: English and Chinese are listed
• Features: the card lists less blur, better faces in medium and long shots, a natural look with less skin shine, fantasy effects and camera control
• Audio: audio references can now help adjust the character's color tone
• License: apache-2.0
• Cost to run: none named, no API key or account. The online demo is on a separate platform
• Data it reaches: it downloads model files from Hugging Face and works on the images and prompts you give it in ComfyUI
• Not for: prompts far longer than about 5,000 characters, since the card recommends staying within that

On Hugging Face, 13 November 2026

• Task: image to video
• Library: minimax-h3
• Downloads in the last 30 days: 1,145,318
• Likes: 986
• License: apache-2.0
• Languages: en, zh
• Based on: MiniMaxAI/MiniMax-H3
• Quantized versions: 5
• Tags: minimax-h3, video-generation, text-to-video, image-to-video, video-to-video, reference-to-video, comfyui, fine-tuned, singularity, int8

Links

• Model: https://huggingface.co/WarmBloodAban/Minimax-h3_Singularity
• Files: https://huggingface.co/WarmBloodAban/Minimax-h3_Singularity/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download WarmBloodAban/Minimax-h3_Singularity

Then follow the usage in its model card. Nothing in this kit runs by itself.