MiniMax H3 Fastest Workflow | 6GB VRAM 16GB RAM — v3.0 — v10.0 (NSFW & SFW_HD)

This MiniMax H3 workflow is for image-to-video, video-to-video, and video upscaling; it loads image or video inputs and includes a SaveVideo output.

Its graph includes MiniMaxH3ReferenceToVideo, RTXVideoSuperResolution, CreateVideo, PreviewImage, and SaveVideo nodes.

Input nodes include LoadImage, LoadAudio, and VHS_LoadVideo; the graph also includes PreviewImage and SaveVideo.

The title includes “6GB VRAM 16GB RAM.”

The listed model files are minimax_h3_audio_vae_fp32.safetensors, minimax_h3_ref2va_pruned_int8_convrot.safetensors, minimax_h3_video_vae_fp16.safetensors, qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors, and taeh3.safetensors.

The listed ComfyUI node packs are comfyui_nvidia_rtx_nodes, comfyui-kjnodes, comfyui-obvpm, comfyui-spectrum-minimax-h3, comfyui-videohelpersuite, and plaguekind-nodes.

Before installing the listed node packs, check which ComfyUI packages correspond to comfyui_nvidia_rtx_nodes and plaguekind-nodes.

Suggestion · not verified

The page’s V11 guidance describes 4, 6, and 8 steps as speed, quality-and-consistency, and highest-quality options.

Suggestion · not verified

The listing title identifies v10.0, while the description begins with a V11 update; check which settings apply to the version you use.

Suggestion · not verified

The title lists 6GB VRAM and 16GB RAM without saying whether those are minimum or recommended requirements; verify hardware fit before using it.

Suggestion · not verified