⚙️ MiniMax H3 Prompt Enhancer — V7.3 Turn simple ideas into detailed, structured prompts for MiniMax video generation.
This ComfyUI workflow is a prompt enhancer built around the official MiniMax H3 prompting guide. It helps transform short, rough ideas into clearer and more effective video prompts — without any complex LLM installation.
Supports T2VA, I2VA, FL2VA, L2VA, and REF2V workflows.
List of models: https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF
Choose the best for your hardware. *MTP models incompatible with the Custom LLM node*
Custom nodes needed (also in comfyui manager):
LLM Text Processor: ⚠️*Must be insalled manually from here, there is a bug for that node in the node manager*⚠️ https://github.com/KingManiya/ComfyUI-LLM-text-processor
Comfyui-RMBG: https://github.com/1038lab/ComfyUI-RMBG
rgthree-comfy: https://github.com/rgthree/rgthree-comfy
✨ What's New in V7
Added a video input node and a second QwenVL node (automatic download on first run for this model) with instructions on how to use.
Backed prompt refining.
Model Change. Now using Qwen3.5-9B-The-Defiant-Fable-Uncnr-Heretic-NEO-MAX-Q8_0. (Download directly in workflow. Extremely accurate, runs fast and easy on 16 VRAM)
V7.1
Separated REF2V Prompt in backend for better accuracy.
Removed time stamps instructions for video reference as it was exceding the video duration most of the time.
V7.2
Prompt cleaning.
V7.3
Prompt cleaning.
Added Number of shots and video duration nodes, no need to input manually now!
🎯 The Goal
This workflow isn’t meant to blindly generate the “perfect” prompt for you. Instead, it gives you a well-structured starting point that you can quickly review, tweak, and customize before sending it to MiniMax.
Simple idea → Enhanced prompt → Your edits → Video generation
💡 Tip: The generated prompt is a starting point. Read through it and adjust the details, actions, camera movement, timing, and style to match your exact vision.
Feel free to use, modify, and integrate this workflow into your own projects. If you create something with it or improve the prompting method, share your results and prompting tips in the comments so the community can keep improving it!
If you use this workflow in your own workflow pack, a quick mention would be greatly appreciated. ❤️
Description
Removed some nodes and replaced with basic comfyui nodes to reduce dependencies.
Fixed a bug where inputing only one image in Expand concept and Ref2v sections would only output a black square.
FAQ
Comments (8)
For those having trouble with LLM node: install manually from link in description as there is a link error in node manager!
i'm so glad you used my suggestion ! :) !
@EvilFear it works like a charm and fast AF! Of course i'm going to use it lol. On a side note, i'm testing with the time stamps in the promter, is it an anoying feature that should be removed? Its all over the place sometimes. Inputing a duration of 10 seconds and it put timestamps up to 18 or 25 seconds :/
@KiraNugget hmm, i personally tweak a little bit your prompt which is very good, i use an other workflow but i wanted to check how you made it work :p (i combined 2 workflows, plus my own tweak ^//^!) when i prompt i always start my prompt manually with FL2VA 5 seconds. the time stamps stay in that range. another thing, i removed <d> - </d>, which i was using before. i was wondering why i was getting weird residual voice, it's apparently a known bug and (take it with a gain of salt, i did not looked up that info) the team behind Minimax H3 say to not use it for the time being. I tested without <d> and i get less problematic voice residual, i side track a little bit, but i wanted to share :)
@EvilFear that is good to know for future updates. thanks a lot! So Just dialogue between " " I guess works best?
@KiraNugget if there's many people in the scene the ID (s1, s2 etc) is super useful, i normal just use (if solo) with a ( what ever you want for the voice) say : 'bla bla bla'. you can even do <sigh>hello<sigh> how are you.., and the 'hello' will be accompanied with a sigh, the model is pretty good to understand the tone of the scene. the 'sigh' pretty damn accurate !
@EvilFear yeah its amazing, in one of my video, the model saw that there was a mirror in the bathroom and created a perfect reflection unprrompted. Best model for sure
@KiraNugget yup ! I'm done with Wan 2.2 :p