seesee75-commits /ComfyUI-MiniMaxH3-Director

About ComfyUI-MiniMaxH3-Director

MiniMax H3 Director drops a timeline editor into ComfyUI for MiniMax H3 video. You drag images, video and music onto tracks, then write one prompt per shot. Prompts compile as you type, so you can read them before you render. A port of WhatDreamsCost's LTX Director, it needs about 60 GB of models.

Best for: Anyone already working in ComfyUI who wants shot-level control over MiniMax H3 video and audio.

• Edits on a timeline with main, reference-video and audio tracks and a seconds ruler
• Shows the compiled prompt, shot count, frame count and word count before you run
• Keeps subject slots @ref1–@ref9 with reference kinds and retention markers
• Adds five nodes, including Retake Stitch and Enhance Prompt
• Regenerates a chosen range with Retake mode and splices it back into the video
• Takes width, height, start, end and duration as connection-only inputs

1. 1
   Install from ComfyUI Manager (search "MiniMax H3 Director"), or clone it: cd ComfyUI/custom_nodes then the command below.

   git clone https://github.com/seesee75-commits/ComfyUI-MiniMaxH3-Director

   Copy
2. 2
   Restart ComfyUI and hard-reload the browser tab (Ctrl+F5). There's nothing to pip install.
3. 3
   Download the MiniMax H3 checkpoints, text encoder and VAEs from Comfy-Org/MiniMax-H3 into the matching ComfyUI/models/ folders, with no subfolders.
4. 4
   Open the command below and fix any red nodes, usually the model dropdowns.

   example_workflows/MiniMax H3 Director.json

   Copy
5. 5
   Double-click a segment on the main track and type the shot. Drag an image onto the track for a first-frame anchor if you want one.
6. 6
   Read the COMPILED PROMPT panel, then Run.