MiniMax H3 Update, No more Waxy Skin, No More Distorted Faces, 8steps=20steps

#comfyui #comfyuitutorial #ai #aivideo #aiimages #minimaxh3 #seedance2 #minimax #comfy #seedance25 #seedance #opensource Resouces: Workflow and App(download or run on cloud): Minimax H3 Cinematic All-In-One Version https://www.runninghub.ai/post/210165... head shot + 3 View character sheet reference worklow https://www.runninghub.ai/post/208892... Official H3 Prompt Skills: https://github.com/MiniMax-AI/MiniMax... Custom Node: https://github.com/LBH-123-AI/Comfyui... Models(other model links are in the workflow note): https://huggingface.co/smhfacct/Minim... https://huggingface.co/TenStrip/Minim... https://huggingface.co/LBH-123-AI/Min... Prompt Making: https://chatgpt.com/s/cx_6aaf3425e338... ComfyUI Installation Guide:    • The Simplest Way to Use ComfyUI   ComfyUI MiniMax H3 just got another upgrade. In this video, I share my updated ComfyUI workflow for H3 using one new speed lora and two cinematic loras, mainly focusing on a few old problems I've been trying to fix: the oily/plastic look from acceleration LoRAs, blurry distant faces, and quality loss in high-motion scenes. I also added a two-stage latent upscale workflow, so you can generate a quick low-res preview first, find a good seed, and then refine it to around 1080p. This saves quite a bit of time compared to doing the full-resolution generation every time. The workflow also uses a merged H3 model, so text-to-video, image-to-video, first/last frame, and multi-reference generation can all be handled in basically the same workflow. I'll go through the settings, LoRAs, sampling steps, and some tips from my own testing. Hope this helps if you're playing with MiniMax H3. More H3 and ComfyUI tests coming soon. Chapters 00:00 H3 Cinematic Showcase 01:18 What's New: Better Skin, Faces & Audio 03:45 ComfyUI Workflow Overview 04:18 Setting Up Image, Audio & Video References 05:13 Merged Models: FL2VA vs. Ref2VA 06:35 One Workflow, Multiple Generation Modes 07:32 The LoRA Sweet Spot: 8 Steps vs. 20 Steps 08:54 Cinema vs. Film LoRA: Details & Distant Faces 10:38 LoRA Settings: Speed, Trigger Words & Weights 11:30 Two-Stage Generation & Latent Upscaling 13:40 Preview with Audio Before Refining 14:48 Sampling Steps: 8 + 4 15:19 First-Pass Testing & Seed Control 16:31 Choosing Your Output Resolution: 1MP vs. 2MP 17:53 Generation Speed & GPU Performance 18:49 Preparing H3 Prompts with the Prompt Skill 19:08 Simplified App Interface & API Access 19:55 Local vs. Cloud GPU 20:38 Creating Videos Longer Than 15 Seconds 21:08 30-Second Demo: Story, References & Prompts 22:17 VRAM Requirements for Longer Videos 22:46 Workflow Recap & Final Thoughts