<p>Tutorial inYouTube：</p><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/3yrxC_qGDi4" start="0"></iframe></div><p>Register through my online workflow to receive 1,000 free RunningHub credits:</p><p><a target="_blank" rel="ugc" href="https://www.runninghub.ai/post/2099399709893853185/?inviteCode=rh-v1497">https://www.runninghub.ai/post/2099399709893853185/?inviteCode=rh-v1497</a></p><p>ComfyUI-Yusu-MiniMaxH3-Unified:</p><p><a target="_blank" rel="ugc" href="https://github.com/yusu-02/ComfyUI-Yusu-MiniMaxH3-Unified">https://github.com/yusu-02/ComfyUI-Yusu-MiniMaxH3-Unified</a></p><p></p><h2 id="runninghub-canvas-version">RunningHub Canvas Version</h2><p><a target="_blank" rel="ugc" href="https://rhtv.runninghub.ai/camp/view/2100066397379215361/?inviteCode=rh-v1497">https://rhtv.runninghub.ai/camp/view/2100066397379215361/?inviteCode=rh-v1497</a></p><h2 id="main-features">Main Features</h2><ul><li><p>MiniMax H3 video generation</p></li><li><p>LightX2V 8-step accelerated workflow</p></li><li><p>Text-to-video generation</p></li><li><p>Image-to-video generation</p></li><li><p>Reference-to-video generation</p></li><li><p>Unified multi-purpose workflow layout</p></li><li><p>Optional cinematic LoRA</p></li><li><p>Simplified prompt and reference-image inputs</p></li><li><p>Adjustable duration and resolution</p></li><li><p>Suitable for product videos, AI UGC ads, character videos, and cinematic experiments</p></li></ul><h2 id="cinematic-lora">Cinematic LoRA</h2><p>The workflow includes a cinematic LoRA stage that can help reduce some of the unnatural facial lighting often seen in AI-generated videos while adding a more grounded cinematic finish.</p><p>I normally use a relatively low LoRA strength of approximately 0.3–0.5. Higher values may change the original image or character appearance too aggressively, so I recommend starting low.</p><h2 id="recommended-workflow">Recommended Workflow</h2><ol><li><p>Select the required generation mode.</p></li><li><p>Enter the video prompt.</p></li><li><p>Upload the reference images in the order specified by the prompt.</p></li><li><p>Confirm which reference is Image 1, Image 2, and so on.</p></li><li><p>Set the duration and output resolution.</p></li><li><p>Run a lower-resolution test first.</p></li><li><p>Review motion, character consistency, hand movement, and prompt logic.</p></li><li><p>Increase the resolution only after the test result is stable.</p></li></ol><p>MiniMax H3 follows prompts closely. This is useful when the instructions are clear, but logical problems in a prompt may also be reproduced very literally. Complex hand movements and detailed physical interactions should be tested carefully.</p><h2 id="notes">Notes</h2><p>This workflow does not remove the need to review prompts and test generated results. Different models have different strengths, and some shots may still require prompt revisions, rerolls, or editing.</p><p>The attached video demonstrates the workflow, real product-video examples, common generation errors, and the practical differences between MiniMax H3 and Seedance.</p>

Files with SHA256: 022c69da5d7d8b5fd458e24b8ca45d60852487079ae7483a752edb0351eaa078