Minimax-LowVRAM — INITIA — PERFECTA

This MiniMax H3 ComfyUI workflow is for video creation on low-VRAM GPUs, with text-to-video and image-to-video paths; it includes image loading and a video-saving step.

Its listed nodes include MiniMaxH3ImageToVideo and CreateVideo.

For image-to-video, it includes LoadImage; for output, it includes SaveVideo.

For the ComfyUI MiniMax H3 files, use the official source at https://huggingface.co/Comfy-Org/MiniMax-H3/tree/main.

Suggestion · non vérifié

The workflow lists these model files: minimax_h3_audio_vae_fp32.safetensors, minimax_h3_fl2va_pruned_int8_convrot.safetensors, minimax_h3_turbo_v4_step600_ema.safetensors, minimax_h3_video_vae_fp16.safetensors, qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors, and taeh3.safetensors.

It also lists these ComfyUI node packs: comfyui_memory_cleanup, comfyui_nvidia_rtx_nodes, comfyui-kjnodes, comfyui-minimax-h3-turbo, and RES4LYF.

For Turbo LoRA, the named sources are larryvrh/MiniMax-H3-Turbo-Lora and Lightx2v/Minimax-h3-Turbo.

For live previews, consider Kijai’s MiniMax-H3-TAE node.

Suggestion · non vérifié

The description targets low-VRAM GPUs but gives no specific GPU model or VRAM figure; check that your card meets your needs.

Suggestion · non vérifié

The workflow lists comfyui_nvidia_rtx_nodes at version 0.1.3; confirm it is available in your ComfyUI setup.

Suggestion · non vérifié