Model reference · open weights
MiniMax-H3-Lora-Pruned-ComfyUI is an open-weight video model from Abiray, listed in the AxForge catalogue. AxForge can bring it up on EU-owned hardware for you on request — with the licence handled where one is required.
About
MiniMax-H3 Turbo LoRA (ComfyUI Pruned) This is a ComfyUI-ready, pruned version of the MiniMax-H3 Turbo LoRA. It is designed to dramatically accelerate both video and audio generation for the MiniMax-H3 model. By using this adapter, you can drastically cut down the required sampling steps while still generating native, synchronized 32kHz stereo audio alongside your video. 🙏 Acknowledgements & Credits Original Trainer & Model Creator: larryvrh/MiniMax-H3-Turbo-Lora 🚀 What Makes This Version Different? The initial release of this LoRA required a standalone custom Python script because standard ComfyUI samplers struggled to handle the dual-schedule (video + audio) mechanics without blowing out the audio track. This specific .safetensors file has been structurally pruned and reformatted so it works seamlessly inside standard ComfyUI workflows. You can load it using a standard Load LoRA node. It is optimized to run alongside the pruned or curve-form MiniMax-H3 checkpoints designed for ComfyUI. 📁 Installation & Setup To get started, simply download LoRA file and place it into your ComfyUI models directory: ComfyUI/models/loras/ Workflow in the repo https://huggingface.co/Abiray/MiniMax-H3-Turbo-Lora-ComfyUI/blob/main/MinimaxH3turboworkflow.json ⚙️ Recommended ComfyUI Settings For the best visual quality and properly synced audio, we highly recommend dialing in the following parameters in your workflow: Base Model: Use the pruned or curve-form MiniMax-H3 ComfyUI checkpoint. Video Sigma Shift: 12 Audio Sigma Shift: 6 Steps: 8 to 12 Sampler: resmultistep LoRA Strength: 0.8 to 1.8 💡 Pro-Tip on Steps vs. Strength: You can generally reduce your sampling steps even further by cranking up the LoRA strength. If you are comparing different models or files, make sure to lock in your seed, prompt, resolution, sampler, and workflow to ensure a fair test. Want to speed up your generation times even more? This workflow has been tested and confirmed to work with the following optional ComfyUI attention and execution accelerators: SageAttention Sol Attention Gradient Spectrum 🔊 Audio Troubleshooting & Mismatches Because MiniMax-H3 synthesizes video and audio simultaneously using two co
Summarised from the published model card. Read the full card on the HuggingFace links below.
Specifications
| Maker | Abiray |
|---|---|
| Type | Video models |
| Variants | 1 |
| Runs with | diffusers |
| Based on | MiniMaxAI/MiniMax-H3 |
| Released | 2026-08-06 |
| Popularity | 21k downloads / month |
| Likes | 48 |
| Licence | Commercial licence needed |
How it works
Variants
Open weights ship in several sizes and precisions. One page, all the variants — pick the one that fits your GPU. VRAM figures are estimates from model size.
| Variant | Params | Precision | VRAM | Fits 16 GB | Weights |
|---|---|---|---|---|---|
| MiniMax-H3-Turbo-Lora-Pruned-ComfyUI | — | BF16 | — | — | Weights ↗ |
Using it via the API
Once AxForge deploys minimax-h3-lora-pruned-comfyui for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (minimax-h3-lora-pruned-comfyui below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/videos/generations \
-H "Authorization: Bearer $AXFORGE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"minimax-h3-lora-pruned-comfyui","prompt":"a drone shot over a forest"}'
Details
Tags
Licence
The weights are open but its licence needs a commercial agreement for business use. AxForge can arrange that licence and host the model for you — you pay AxForge, we settle with the model’s maker. Ask us for a quote. Read the licence ↗
Explore