The Workflow was setup to have a clean "GUI" showing only parameters that matter, so you might want to toggle off Link visibility.

V1.0 Minimax H3: IMAGE or TEXT or REFERENCE to Video with Ollama

Workflow features:

• single Image, First Frame/Last Frame, Text or Reference to VIDEO

• can use up to 4 images, 1 video, 1 audio input as Reference to generate videos

• uses Ollama with dedicated system prompts to enhance simple user prompts

• applies RTX Video Super Resolution to upscale to a final resolution very fast

• can toggle the following accelerators: Turbo Lora, Sage Att., EasyCache, Spectrum, FirstBlockCache, Sol. Att.

Downloads:

• Models (IT2V and REF2V, int8_convrot recommended): https://huggingface.co/Comfy-Org/MiniMax-H3/tree/main/diffusion_models

• Textencoder (nvfp4) : https://huggingface.co/Comfy-Org/MiniMax-H3/tree/main/text_encoders

• VAEs (audio & video) : https://huggingface.co/Comfy-Org/MiniMax-H3/tree/main/vae

• Turbo Lora (resized): https://huggingface.co/Kijai/MiniMax-H3_comfy/tree/main/loras

  • LightX2V Turbo Loras: https://huggingface.co/lightx2v/Minimax-h3-Turbo/tree/main

• Ollama model (tested w/ 26b): https://ollama.com/huihui_ai/gemma-4-abliterated or any other model ideally with vision capabilities, like (9b) :https://ollama.com/huihui_ai/qwen3.5-abliterated

Save locations:

📂 ComfyUI/
├── 📂 models/
│   ├── 📂 vae/
│   │   ├── minimax_h3_video_vae_fp16.safetensors
│   │   └── minimax_h3_audio_vae_fp32.safetensors
│   ├── 📂 diffusion_models/
│   │   └── minimax_h3_fl2va_pruned_int8_convrot.safetensors
│   └── 📂 text_encoders/
│       └── qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors

Ollama help:

1. Install Ollama from https://ollama.com/

2. download a model: Go to a model page, chose a model , then hit the copy button, i.e. https://ollama.com/huihui_ai/qwen3-vl-abliterated

3. open terminal and paste the model name, i.e.: ollama run huihui_ai/qwen3-vl-abliterated

4. model will be downloaded and can be selected in green comfy node "Ollama Connectivity". Hit "Reconnect" to refresh.

Custom Nodes used:

• https://github.com/rgthree/rgthree-comfy

• https://github.com/kijai/ComfyUI-KJNodes

• https://github.com/yolain/ComfyUI-Easy-Use

• https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite

• https://github.com/stavsap/comfyui-ollama

• https://github.com/Comfy-Org/Nvidia_RTX_Nodes_ComfyUI

• https://github.com/xmarre/ComfyUI-Spectrum-MiniMax-H3

• https://github.com/duckyshell/ComfyUI-MiniMaxH3-FirstBlockCache