<h1 id="minimax-h3-turbo-lora">⚡ MINIMAX H3 TURBO LoRA</h1><h2 id="4-step-audio-+-video-acceleration">4-STEP AUDIO + VIDEO ACCELERATION</h2><blockquote><p><strong>EXTREME SPEED // NATIVE AUDIO // COMFYUI READY</strong></p></blockquote><p>MiniMax H3 Turbo LoRA is an experimental acceleration adapter designed to drastically reduce MiniMax H3 video generation time while preserving its native synchronized stereo audio capabilities.</p><p>It reduces sampling from the usual higher step count down to only <strong>4 steps</strong>, providing extremely fast previews and significantly shorter generation times.</p><p>The LoRA supports both:</p><pre><code>OFFICIAL EKKIVOK WORKFLOW :</code></pre><p><a target="_blank" rel="ugc" href="https://civarchive.com/models/2838258/minimax-h3-4-steps-turbo-video-aio-workflow">https://civarchive.com/models/2838258/minimax-h3-4-steps-turbo-video-aio-workflow</a></p><pre><code>TEXT TO VIDEO
IMAGE TO VIDEO
FIRST / LAST FRAME VIDEO</code></pre><hr /><h1 id="early-experimental-release">⚠️ EARLY EXPERIMENTAL RELEASE</h1><p>This LoRA is currently an <strong>experimental and under-trained prototype</strong>.</p><p>The 4-step mode provides impressive speed, but it may produce:</p><ul><li><p>Softer image quality</p></li><li><p>Reduced fine details</p></li><li><p>Unstable textures</p></li><li><p>Motion ghosting</p></li><li><p>Motion trails</p></li><li><p>Lower overall quality compared with the full MiniMax H3 model</p></li></ul><p>For improved visual quality, using <strong>6 to 8 steps</strong> may produce cleaner results while still remaining much faster than standard generation.</p><hr /><h1 id="recommended-settings">🚀 RECOMMENDED SETTINGS</h1><pre><code></code></pre><pre><code>LoRA Strength : 1.0
Steps         : 6–8 recommended
Fast Preview  : 4 steps
Scheduler     : simple
CFG           : 1.0 or workflow default
Frame Rate    : 24 FPS
Resolution    : multiples of 32
Short Edge    : approximately 768 px</code></pre><h3 id="recommended-lora-file">Recommended LoRA file</h3><pre><code></code></pre><pre><code>minimax_h3_turbo_4step_ckpt500.safetensors</code></pre><p>The non-EMA <code>ckpt500</code> version is currently the recommended checkpoint.</p><p>The EMA version may produce softer results and can be more prone to ghosting or visible motion trails.</p><hr /><h1 id="native-synchronized-audio">🔊 NATIVE SYNCHRONIZED AUDIO</h1><p>MiniMax H3 generates video and audio together:</p><pre><code></code></pre><pre><code>VIDEO STREAM
+
32 kHz STEREO AUDIO</code></pre><p>The audio is generated natively and synchronized directly with the video.</p><p>A dedicated Turbo sampler is required because MiniMax H3 uses different sampling trajectories for video and audio:</p><pre><code></code></pre><pre><code>Video Flow Shift : 12
Audio Flow Shift : 3</code></pre><p>Using a standard ComfyUI sampler at only 4 steps may incorrectly sample the audio component, resulting in distorted, saturated, or unusable sound.</p><hr /><h1 id="required-comfyui-custom-node">🧩 REQUIRED COMFYUI CUSTOM NODE</h1><p>Install the following custom node through ComfyUI Manager:</p><pre><code></code></pre><pre><code>ComfyUI-MiniMax-H3-Turbo</code></pre><p>Official repository:</p><p><a target="_blank" rel="ugc" href="https://github.com/Larryvrh/ComfyUI-MiniMax-H3-Turbo">https://github.com/Larryvrh/ComfyUI-MiniMax-H3-Turbo</a></p><p>The node pack includes:</p><pre><code></code></pre><pre><code>MiniMax-H3 Turbo LoRA
MiniMax-H3 Turbo Sampler</code></pre><p>The custom node also supports pruned MiniMax H3 checkpoints by automatically applying the required AdaLN LoRA layers during generation.</p><hr /><h1 id="base-model-compatibility">🧠 BASE MODEL COMPATIBILITY</h1><p>The Turbo LoRA is compatible with several MiniMax H3 model formats:</p><pre><code></code></pre><pre><code>BF16
INT8 ConvRot
Pruned INT8
Pruned FP8</code></pre><p>The dedicated Turbo node automatically detects the model structure and applies the LoRA appropriately.</p><hr /><h1 id="installation">📂 INSTALLATION</h1><p>Place the LoRA file inside:</p><pre><code></code></pre><pre><code>ComfyUI/models/loras/</code></pre><p>Recommended workflow structure:</p><pre><code></code></pre><pre><code>Load Diffusion Model
        ↓
MiniMax-H3 Turbo LoRA
        ↓
MiniMax-H3 Turbo Sampler
        ↓
Video and Audio Output</code></pre><p>Recommended sampling configuration:</p><pre><code></code></pre><pre><code>Scheduler : simple
Steps     : 4–8</code></pre><p>Do not use a standard sampler for the 4-step audio workflow.</p><hr /><h1 id="original-creator">👤 ORIGINAL CREATOR</h1><h2 id="larryvrh">Larryvrh</h2><p>This LoRA was originally created and trained by <strong>Larryvrh</strong>.</p><h3 id="original-hugging-face-repository">Original Hugging Face repository</h3><p><a target="_blank" rel="ugc" href="https://huggingface.co/larryvrh/MiniMax-H3-Turbo-Lora">https://huggingface.co/larryvrh/MiniMax-H3-Turbo-Lora</a></p><h3 id="official-comfyui-custom-node">Official ComfyUI custom node</h3><p><a target="_blank" rel="ugc" href="https://github.com/Larryvrh/ComfyUI-MiniMax-H3-Turbo">https://github.com/Larryvrh/ComfyUI-MiniMax-H3-Turbo</a></p><p>All credit for the LoRA training, development, research, and dedicated Turbo sampler belongs to <strong>Larryvrh</strong>.</p><hr /><h1 id="civitai-upload-information">📤 CIVITAI UPLOAD INFORMATION</h1><pre><code></code></pre><pre><code>Original Creator : Larryvrh
Civitai Uploader : EKKIVOK</code></pre><p>This Civitai page is a community re-upload intended to make the model easier to discover, download, and use with ComfyUI.</p><p><strong>I am not the original creator of this LoRA.</strong></p><p>Please support and follow the original creator through the official repositories linked above.</p><hr /><h1 id="speed-and-quality-guide">⚡ SPEED AND QUALITY GUIDE</h1><pre><code></code></pre><pre><code>4 STEPS = MAXIMUM SPEED / PREVIEW QUALITY
6 STEPS = BALANCED TURBO MODE
8 STEPS = BEST CURRENT TURBO QUALITY</code></pre><p>The Turbo LoRA is especially useful for:</p><ul><li><p>Rapid prompt testing</p></li><li><p>Motion previews</p></li><li><p>Seed comparison</p></li><li><p>Fast workflow testing</p></li><li><p>Video drafts</p></li><li><p>Quickly testing image-to-video inputs</p></li><li><p>Reducing MiniMax H3 sampling time</p></li></ul><p>For maximum final quality, the standard MiniMax H3 model without the Turbo LoRA and with a higher step count may still produce better results.</p><hr /><blockquote><p><strong>CREATED BY LARRYVRH</strong><br /><strong>UPLOADED TO CIVITAI BY EKKIVOK</strong><br /><strong>MINIMAX H3 // TURBO MODE // FUTURE IN MOTION</strong></p></blockquote><p></p>

Files with SHA256: 82d0acff583b04ad9a4238a7440b584b56094bfb7c4fdb2981f67c7a4784b62d