<h2 id="comfyui-usage-and-node-requirements">ComfyUI Usage &amp; Node Requirements</h2>To run these GGUF models in ComfyUI, install the following custom node packages:<ol><li>GGUF Loader Node :<ul><li><a target="_blank" rel="ugc" href="https://github.com/molbal/ComfyUI-GGUF">ComfyUI-GGUF (GitHub)</a> or install via ComfyUI Node Registry: <a target="_blank" rel="ugc" href="https://registry.comfy.org/publishers/molbal/nodes/comfyui-gguf-reboot">comfyui-gguf-reboot</a></li><li>Used to load the quantized <code>.gguf</code> UNet/diffusion models.</li></ul></li><li>MiniMax H3 Turbo Sampler:<ul><li><a target="_blank" rel="ugc" href="https://github.com/Larryvrh/ComfyUI-MiniMax-H3-Turbo">ComfyUI-MiniMax-H3-Turbo (GitHub)</a> or install via ComfyUI Node Registry: <a target="_blank" rel="ugc" href="https://registry.comfy.org/nodes/comfyui-minimax-h3-turbo">comfyui-minimax-h3-turbo</a></li><li>Provides the dedicated Turbo sampling schedules and shift calculations for 4-step generation.</li></ul></li></ol><h3 id="model-placement">Model Placement</h3>Place the downloaded GGUF files in your ComfyUI models directory:<pre><code>ComfyUI/models/diffusion_models/minimax_h3_fl2v_turbo_4step_v1.0_768p_Q8_CR.gguf</code></pre><h2 id="shift-and-sampling-schedule">Shift &amp; Sampling Schedule</h2><ul><li>Recommended inference step count: 4 steps.</li><li>Minimax H3 Video shift: 6, Audio shift: 3</li><li>Minimax H3 Turbo Sampler</li><li>simple scheduler</li></ul><h2 id="supported-input-modes-(fl2va)">Supported Input Modes (FL2VA)</h2><ul><li>Text-to-Audio-Video (T2VA): No image input connected.</li><li>Image-to-Audio-Video (I2VA): One image connected to <code>first_frame</code>.</li><li>First-and-Last-Frame-to-Audio-Video (FL2VA): Two images connected to <code>first_frame</code> and <code>last_frame</code>.</li></ul><h2 id="comfyui-workflows">ComfyUI Workflows</h2>Example workflows are provided<ul><li><code>workflows/minimax_h3_t2v-turbo-gguf.json</code> — Text-to-Audio-Video (T2VA) Turbo workflow.</li><li><code>workflows/minimax_h3_i2v-turbo-gguf.json</code> — Image-to-Audio-Video (I2VA / FL2VA) Turbo workflow.</li></ul><h2 id="license">License</h2>MiniMax H3 is released under the <a target="_blank" rel="ugc" href="https://huggingface.co/MiniMaxAI/MiniMax-H3/blob/main/LICENSE">MiniMax H3 Community License Agreement</a>.

Files with SHA256: e254c1e576616f0a62041a08f5f325468c5f5e5cfd5cc90cb53e2f2e45bc7906