A quick Minimax H3 news round-up - 14th October 2026
retained text
Another quick Minimax H3 news and goodies round-up, for those who may have missed some items. -> ComfyUI Portable is now officially at version 0.40.0. There's a new official 'Camera Angle' node, and... "MiniMax-H3 single images: Single images are now decoded as the first latent of a clip" and thus can avoid banding. Hopefully this change doesn't bjork the excellent ComfyUI-Fizgig-H3-Still. https://github.com/Comfy-Org/ComfyUI/releases https://github.com/shootthesound/ComfyUI-Fizgig-H3-Still -> ComfyUI NVIDIA RTX VSR Long Video. NVIDIA's excellent quick RTX upscaling, here apparently specifically tailored to upscale long videos. https://github.com/tritant/ComfyUI_NVIDIA_RTX_VSR_Long_Video -> A small collection of action/motion related LoRA sliders. Inertia & Momentum; Movement Amplitude; Progressive Acceleration; Rapid Motion Changes; Motion Speed. https://civitai.com/collections/18270307 -> Advice and a prompt to help maintain visual style with Minimax H3. Triple-click to select/copy the prompt. https://old.reddit.com/r/StableDiffusion/comments/1x3f6x0/how_do_you_maintain_style_in_minimax_h3_ref_mode/ -> Some Reddit advice on improving character voices, and regional dialect / national accents. https://old.reddit.com/r/StableDiffusion/comments/1x4pxv6/is_there_a_way_to_train_minimax_h3_lora_for_new/ https://old.reddit.com/r/StableDiffusion/comments/1x5f6yx/how_to_improve_character_voices_mm_h3/ -> ComfyUI H3 AudioFirst X2 Stream. "The voice is finished first in an 8-step audio-only pass on a tiny 32 × 32px canvas, then locked" for the subsequent video generation. https://github.com/sepiablue-ai/ComfyUI-H3-AudioFirst-X2-Stream -> The new PerfectLab Studio 1.0.3 for Z-Image, Flux, Qwen-Image and SDXL. Potentially useful for generating a large consistent photoshoot set, and then selecting suitable start / middle / end frames from it for subsequent Minimax H3 animation. https://github.com/perfectmodelslab/ComfyUI-PerfectLab -> And finally, the Comfyui-obvpm-timeline (for Comfyui-obvpm) has updated. Includes a new workflow, with *taeh3.safetensors* now providing the live generation preview. https://github.com/chanon/comfyui-obvpm-timeline
document fetched 2026-10-15 00:45 · canonical https://reddit.com/r/comfyui/comments/1x5puo0/a_quick_minimax_h3_news_roundup_14th_october_2026