Minimax-LowVRAM — INITIA — TURBIA

This ComfyUI workflow is for creating MiniMax H3 videos on low-VRAM GPUs: it takes an image, generates video from it, and includes steps to preview and save the result.

Its 23-node graph loads and resizes an image, passes it through MiniMax H3 image-to-video and Turbo sampler nodes, then creates, previews, and saves video.

The workflow uses an image input through `LoadImage`; `PreviewImage` and `SaveVideo` provide preview and video-saving steps.

The download is `minimaxLowvram_turbia.zip`, which contains the workflow file `video_minimax_low_vram_turbo.json`.

Listed model files include `minimax_h3_audio_vae_fp32.safetensors`, `minimax_h3_video_vae_fp16.safetensors`, `qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors`, and `minimax_h3_fl2va_pruned_int8_convrot.safetensors`.

Two other listed model files are `minimax_h3_turbo_4step_ckpt500.safetensors` and `taeh3.safetensors`.

The workflow lists `comfyui_nvidia_rtx_nodes` 0.1.3, `comfyui-kjnodes` 1.4.8, and `comfyui-minimax-h3-turbo` 1.0.0.

The archive contains no model weight files. Arrange the six listed model files separately; source links are listed for the audio VAE, video VAE, and text encoder only.

Vorschlag

Before loading the workflow, check that your ComfyUI setup has the three listed node packs available at the stated versions.

Vorschlag