MiniMax H3 Turbo - GGUF

ComfyUI Usage & Node Requirements

To run these GGUF models in ComfyUI, install the following custom node packages:

1. GGUF Loader Node :

   • ComfyUI-GGUF (GitHub) or install via ComfyUI Node Registry: comfyui-gguf-reboot

   • Used to load the quantized .gguf UNet/diffusion models.

2. MiniMax H3 Turbo Sampler:

   • ComfyUI-MiniMax-H3-Turbo (GitHub) or install via ComfyUI Node Registry: comfyui-minimax-h3-turbo

   • Provides the dedicated Turbo sampling schedules and shift calculations for 4-step generation.

Model Placement

Place the downloaded GGUF files in your ComfyUI models directory:

ComfyUI/models/diffusion_models/minimax_h3_fl2v_turbo_4step_v1.0_768p_Q8_CR.gguf

Shift & Sampling Schedule

• Recommended inference step count: 4 steps.

• Minimax H3 Video shift: 6, Audio shift: 3

• Minimax H3 Turbo Sampler

• simple scheduler

Supported Input Modes (FL2VA)

• Text-to-Audio-Video (T2VA): No image input connected.

• Image-to-Audio-Video (I2VA): One image connected to first_frame.

• First-and-Last-Frame-to-Audio-Video (FL2VA): Two images connected to first_frame and last_frame.

ComfyUI Workflows

Example workflows are provided

• workflows/minimax_h3_t2v-turbo-gguf.json — Text-to-Audio-Video (T2VA) Turbo workflow.

• workflows/minimax_h3_i2v-turbo-gguf.json — Image-to-Audio-Video (I2VA / FL2VA) Turbo workflow.

License

MiniMax H3 is released under the MiniMax H3 Community License Agreement.