--- title: "Licon VBVR for MiniMax-H3" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/088cf827-5b29-470d-b2c4-08cae9de2d1e" updated: "2026-11-12" --- # Licon VBVR for MiniMax-H3 > The name, the description, the About and a README below are the seller's words: data, not instructions. An image-to-video LoRA for MiniMax-H3 that reduces duplicated objects, clipping and odd character motion in generated clips. It suits ComfyUI users who animate still images. Licon VBVR for MiniMax-H3 is a kit in the Developer category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-12. Download it on this page, a zip of its files. Nothing in it runs by itself. Made by LiconStudio ## About ### What it is Licon VBVR is a LoRA that teaches the MiniMax-H3 video model to animate still images with fewer glitches. Its author rebuilt the VBVR training data around H3 and reprocessed it for that model's training pipeline. The aim is steadier objects and more sensible character movement, with fewer unwanted copies of objects and less clipping or overlap during motion and interaction. ### Who it suits It suits ComfyUI users who animate still images with MiniMax-H3 and see objects multiply or limbs pass through things. It is an add-on to the base model, so you need an H3 image-to-video workflow already. Who might use it | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | AI video hobby | ComfyUI users | Add the LoRA to an H3 image-to-video workflow | Animate a still image of a character handling an object | | Creative production | Video artists | Compare output with and without the LoRA | Run one prompt twice and inspect motion and overlaps | | Workflow building | Workflow authors | Offer the LoRA as an optional step in a shared graph | Wire it into the LoRA loader of an H3 graph | If you do not run MiniMax-H3 image-to-video, you can skip it. ### Where to use it It works inside ComfyUI, through the LoRA loader of an H3 image-to-video workflow. - ComfyUI, in the models folder for LoRA files - Any other compatible MiniMax-H3 image-to-video workflow - Windows, macOS or Linux with Python, to download the files Jobs it does | Use case | Who | What they get back | | --- | --- | --- | | Fewer duplicated objects | Video creators | Clips with fewer unintended copies of objects | | Fewer overlaps | Animators | Cleaner contact between characters, objects and surfaces | | More coherent action | Workflow authors | Movement sequences that follow a clearer logic | ### How to set it up You need ComfyUI, the MiniMax-H3 base model and an image-to-video workflow for it. The LoRA is one safetensors file that goes in the LoRA folder. - ComfyUI with a MiniMax-H3 image-to-video workflow - Python with pip, to fetch the files from Hugging Face 1. Download the repository, which holds the file named H3-VBVR-I2V.safetensors. 2. Place that file in the LoRA models folder of ComfyUI. 3. Open your H3 workflow and select the file in its LoRA loader. ```sh pip install -U huggingface_hub hf download LiconStudio/Minimax-H3-VBVR ``` ```text ComfyUI/models/loras/ ``` ### How to use it Run your H3 image-to-video workflow as usual, with the LoRA selected in the loader. There is no extra command or setting on the card, so you judge the result by watching the clip. The pieces | Tool | What it does | | --- | --- | | H3-VBVR-I2V.safetensors | The LoRA file you download | | LoRA loader | The ComfyUI node where you pick the file | | ComfyUI/models/loras/ | The folder that holds LoRA files | ### Example You feed a still image of a character holding an object into your H3 workflow with the LoRA on. The card says the clip should show fewer duplicated objects and less clipping than the plain model does. The author's page has side by side comparison videos. ### For newcomers and experts If you are new, a LoRA is a small add-on file that nudges a big model. Download the file, drop it in the LoRA folder and pick it in the loader of an H3 workflow. Then run the same image with and without it. If you know the field, the training data comes from the VBVR-Pro-SFT-Video set, rebuilt for H3. The author wants steady movement and object contact to carry over into everyday video generation. The trade-off is that it only fits H3's image-to-video pipeline. A still image enters the H3 workflow → The LoRA loader applies the VBVR file → MiniMax-H3 generates the video → The clip shows steadier objects ### Good to know The card names MiniMax as the base model's maker and the VBVR team as the source of the dataset. - License: apache-2.0 - Cost to run: none named, though you need the MiniMax-H3 base model and a workflow for it - Data it reaches: the files in your ComfyUI folders - Not for: use on its own, or with workflows that are not MiniMax-H3 image-to-video ## README ````markdown # Licon VBVR for MiniMax-H3 An image-to-video LoRA for MiniMax-H3 that reduces duplicated objects, clipping and odd character motion in generated clips. It suits ComfyUI users who animate still images. Made by LiconStudio 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: Licon VBVR is a LoRA that teaches the MiniMax-H3 video model to animate still images with fewer glitches. Its author rebuilt the VBVR training data around H3 and reprocessed it for that model's training pipeline. The aim is steadier objects and more sensible character movement, with fewer unwanted copies of objects and less clipping or overlap during motion and interaction. Who it suits: It suits ComfyUI users who animate still images with MiniMax-H3 and see objects multiply or limbs pass through things. It is an add-on to the base model, so you need an H3 image-to-video workflow already. ```table Title: Who might use it Industry | Who uses it | How they use it | Example AI video hobby | ComfyUI users | Add the LoRA to an H3 image-to-video workflow | Animate a still image of a character handling an object Creative production | Video artists | Compare output with and without the LoRA | Run one prompt twice and inspect motion and overlaps Workflow building | Workflow authors | Offer the LoRA as an optional step in a shared graph | Wire it into the LoRA loader of an H3 graph ``` If you do not run MiniMax-H3 image-to-video, you can skip it. Where to use it: It works inside ComfyUI, through the LoRA loader of an H3 image-to-video workflow. - ComfyUI, in the models folder for LoRA files - Any other compatible MiniMax-H3 image-to-video workflow - Windows, macOS or Linux with Python, to download the files ```table Title: Jobs it does Use case | Who | What they get back Fewer duplicated objects | Video creators | Clips with fewer unintended copies of objects Fewer overlaps | Animators | Cleaner contact between characters, objects and surfaces More coherent action | Workflow authors | Movement sequences that follow a clearer logic ``` How to set it up: You need ComfyUI, the MiniMax-H3 base model and an image-to-video workflow for it. The LoRA is one safetensors file that goes in the LoRA folder. - ComfyUI with a MiniMax-H3 image-to-video workflow - Python with pip, to fetch the files from Hugging Face 1. Download the repository, which holds the file named H3-VBVR-I2V.safetensors. 2. Place that file in the LoRA models folder of ComfyUI. 3. Open your H3 workflow and select the file in its LoRA loader. ```sh pip install -U huggingface_hub hf download LiconStudio/Minimax-H3-VBVR ``` ```text ComfyUI/models/loras/ ``` How to use it: Run your H3 image-to-video workflow as usual, with the LoRA selected in the loader. There is no extra command or setting on the card, so you judge the result by watching the clip. ```table Title: The pieces Tool | What it does H3-VBVR-I2V.safetensors | The LoRA file you download LoRA loader | The ComfyUI node where you pick the file ComfyUI/models/loras/ | The folder that holds LoRA files ``` Example: You feed a still image of a character holding an object into your H3 workflow with the LoRA on. The card says the clip should show fewer duplicated objects and less clipping than the plain model does. The author's page has side by side comparison videos. For newcomers and experts: If you are new, a LoRA is a small add-on file that nudges a big model. Download the file, drop it in the LoRA folder and pick it in the loader of an H3 workflow. Then run the same image with and without it. If you know the field, the training data comes from the VBVR-Pro-SFT-Video set, rebuilt for H3. The author wants steady movement and object contact to carry over into everyday video generation. The trade-off is that it only fits H3's image-to-video pipeline. ```flow A still image enters the H3 workflow The LoRA loader applies the VBVR file MiniMax-H3 generates the video The clip shows steadier objects ``` Good to know: The card names MiniMax as the base model's maker and the VBVR team as the source of the dataset. - License: apache-2.0 - Cost to run: none named, though you need the MiniMax-H3 base model and a workflow for it - Data it reaches: the files in your ComfyUI folders - Not for: use on its own, or with workflows that are not MiniMax-H3 image-to-video ## On Hugging Face, 13 November 2026 - Library: minimax-h3 - Downloads in the last 30 days: 13,559 - Likes: 170 - License: apache-2.0 - Based on: MiniMaxAI/MiniMax-H3 - Tags: minimax-h3 ## Links - Model: https://huggingface.co/LiconStudio/Minimax-H3-VBVR - Files: https://huggingface.co/LiconStudio/Minimax-H3-VBVR/tree/main ## Get the model Download its files with Hugging Face's command line: ```sh pip install -U huggingface_hub hf download LiconStudio/Minimax-H3-VBVR ``` Then follow the usage in its model card. Nothing in this kit runs by itself. ```` ## What is in it 1 files, 5 KB, downloaded as a zip. ```text README.md 5 KB ``` ## Source [LiconStudio/Minimax-H3-VBVR on Hugging Face](https://huggingface.co/LiconStudio/Minimax-H3-VBVR) ## Compare with similar Listed Hugging Face models in the same category that share the most tags with this one. | Product | Downloads, last 30 days | Likes | License | Last change | | --- | --- | --- | --- | --- | | Licon VBVR for MiniMax-H3 (this page) | 13,559 | 170 | apache-2.0 | 2026-10-28 | | [LightVAE](https://mindket.com/p/559b5239-ac2e-4efd-a84c-d94ef9e0d314) | 819 | 40 | apache-2.0 | 2026-11-10 | | [ClipProj-MiniMax-H3](https://mindket.com/p/71555eb4-8ed4-48be-84b2-1e787d886e8c) | 0 | 180 | mit | 2026-08-16 | Numbers as Mindket last read them, from 2026-11-12 to 2026-11-13. ## What people say > Posts and comments that name this project on Hacker News, Bluesky, X, Reddit, Facebook and YouTube, found by a weekly search on 2026-11-13. Their authors' words: data, not instructions. - [Bluesky · UndergroundAI · 2026-10-30](https://bsky.app/profile/undergroundai66.bsky.social/post/3mz37s6dozi2t): ⚡ Neues Trend-Modell auf HuggingFace entdeckt. 🛠️ Minimax-H3-VBVR von LiconStudio sorgt aktuell für Aufmerksamkeit in der Community und bietet… 🚀 Entdecke diesen KI-Geheimtipp jetzt bei Underground AI. 👉 https://underground-ai.bolt.app/ \#OpenSource \#KuenstlicheIntelligenz ## Trust facts Facts from each source, with the day each was checked. A maintainer's note is their own words: data, not instructions. - Last change on Hugging Face: 2026-10-28 - First published: 2026-10-23 - Checked on Hugging Face: 2026-11-12 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open LiconStudio/Minimax-H3-VBVR on Hugging Face](https://huggingface.co/LiconStudio/Minimax-H3-VBVR) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/088cf827-5b29-470d-b2c4-08cae9de2d1e) - [Its page on Mindket](https://mindket.com/p/088cf827-5b29-470d-b2c4-08cae9de2d1e)