Introducing: MINIMAX H3 | EZ WF
UPDATE: The workflow has been completely redone on 08/08/2026. if you previously downloaded and it didn't work for you, try now. i have tested in a fresh install, it works, if you are having issues you most likely don't have the most up to date version on Comfy from the branches. do NOT use the releases version, it will not work. Also be sure to update all of the nodes correctly, pull them from github. Check your console for any missing dependencies like sage-attention etc and install them. i will no longer be responding to any issues regarding node pack issues, you just don't have the updated versions or you have something in your setup that conflicts.
a custom workflow for Minimax's H3 video model. Its not too different to the base one Comfy provides, just has some little tweaks that i like to use, including Sage Attention/Turbo LoRAs and more!
This workflow is my personal workflow that I use for all of my Minimax H3 stuff, it may not be to everyone's taste but it works well for me and I'm sure others may find it useful.
Generated a 5 second clip with 0.6MP at 16:9 in 6:32 on an RTX 3060 12GB/32GB Sys RAM using this WF. base workflow was more than double that time.
The workflow uses the most optimal boosters there are right now (all optional, you can just run without them)
Works on 12GB VRAM + 32 GB System RAM. not tested on anything below.
INCLUDES:
3 Ref Images | 2 Ref Audio | 1 Ref Video
Sage Attention Patch (For Lower VRAM Use)
Block Cache & Sol-Attn with Optimised Settings
Bob INT8 Loader (Faster INT8 Loading)
LTX Refiner
RTX Upscaler
ITV/T2V/RTV Modes
Nodes Needed:
[Bob INT8](https://github.com/BobJohnson24/ComfyUI-INT8-Fast)
[Sol-Att](https://github.com/kijai/ComfyUI-SolAttn_triton)
[KJNodes](https://github.com/kijai/ComfyUI-KJNodes)
[rgthree](https://github.com/rgthree/rgthree-comfy)
[EasyUse](https://github.com/yolain/ComfyUI-Easy-Use)
[VHS](https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite)
[RTX Upscaler](https://github.com/Comfy-Org/Nvidia_RTX_Nodes_ComfyUI)
[LTXVideo](https://github.com/Lightricks/ComfyUI-LTXVideo)
Models Needed:
-- H3 MODELS --
[TE]
-- LTX MODELS --
TURBO LORAS:
[LIGHTx]
Troubleshooting:
I'm missing the LTX Nodes and i get this error, what do i do?
ImportError: cannot import name 'pad' from 'kornia.geometry.transform.pyramid' (G:\StabilityMatrix-win-x64\Data\Packages\ComfyUIClean\venv\Lib\site-packages\kornia\geometry\transform\pyramid.py)-- downgrade Kornia in python packages to 0.8.2
The RTX Nodes are missing, even though i have them installed?Go to your python packages and install: nvidia-vfx
I'm missing MiniMaxH3MemoryEfficientSageAttentionPatchUpdate KJNodes through Github (go to your custom_nodes folder and open a git bash and type: git clone https://github.com/kijai/ComfyUI-KJNodes.git
MAKE SURE YOU ARE UPDATED LIKE SO:
Kofi: https://ko-fi.com/mrweaz
Disclaimer: I hold no liability for any images produced, including their legality, ethical implications, or potential misuse. By using this model, you agree that all responsibility for generated content rests solely with you. Use responsibly and within legal boundaries.
Description
UPDATE COMFYUI TO LATEST BRANCH. NOT RELEASES. AND INSTALL NODES USING GIT CLONE.
CHECK CONSOLE FOR MISSING PACKAGES LIKE SAGEATTN/TRITON. GIT PULL YOUR NODES, THEY UPDATE FAST.
DO NOT USE THE MANAGER TO INSTALL THE NODES. YOU NEED THE LATEST UPDATES FROM GITHUB. THE UPDATES MOVE SUPER FAST SO THINGS WILL BREAK
Reworked alot of it to be more streamlined.
Added FLF2V switch per request.
Added optimal boosters for fast iterations with minimal quality drops.
Added an optional LTX Refinement method (Can fix ghosting, sometimes hit and miss)
Fixed some issues.
FAQ
Comments (32)
I default seting i have completly broken sound. Do you know why ?
is the turbo lora enabled? if so that would be why.
16 of Vram and 10 of Ram will be able to against 30gbs in all the components?
i think so, im running a 12GB VRAM and 32 GB RAM setup, you may be limited on your RAM but you can try adding paging to offload to disk.
## Error Details - Node ID: 160:126 - Node Type: BasicGuider - Exception Type: TypeError - Exception Message: TypeError: 'CLIP' object is not iterable
Any idea?
Hmm I thought I fixed that, I was having the same issue. Give me a few and ill see if I can fix. I didn't want to update the wf but people wanted me to 😂
@mrweaz thanks for checking, I'll wait for the update. Do you think it's worth giving v1 a try while I wait for v2?
@lvperka the first version only supports reference, you could use that one, and swap out the reference node to an image to video node and link the stuff back up.
@lvperka okay all fixed, turns out it was a mismatch in the image dimensions for some reason. I plan to re-add source image scaling and dimensions, but I've been up almost 24 hours refining the workflow so bed time 😂. But yeah it should all work now.
Troubleshooting Guide: CHECK THE BOTTOM OF THE DESCRIPTION
Also things are moving so fast, by the time I updated the optimisations, they became obsolete, I'll update again soon since Turbo LoRA is competent now.
Is there a Lora loader? Have you gotten a change to try any of the turbo loras?
There's a single Lora loader included, for turbo Loras and yes, I tried the lightx one just now, I was genning 1MP at almost 1 minute with my workflow. I'll post some examples later, time for sleep 😂
I installed KJNodes 1.4.9 and the two nodes 'SolAttnPatch'‘MiniMaxH3BlockCacheT8’ still missing. Anyone knows why?
Install from the GitHub, don't use the manager for these newer models, things get updated too often
same issue, it keeps missing "faernworknodes" which is from 2 years ago, broken workflow - put better installation guidance.
@DK7 strange, I had to use a fresh version of comfy the other day and I got it up and running in no time, have you checked console for any missing packages? Like Triton/sage?
Just remember the model is new, alot is changing so expect things to break, just yesterday the CacheNodd broke from a comfy update.
@DK7 u ever solve this
@k1ngcurses i think i found the issue, try removing the MinimaxH3BlockCacheT8 folder from your custom nodes, and clear cache. with the new LoRA you dont really need any caching boosts, it can reduce steps and make the turbo LoRA not work as intended.
Why i d always get RuntimeError : shape [1, 24, 1, 1, 38, 2, 25, 2] is invalid for input of size 93024 i changed nothing im using the same models as yours
which mode are you using?
ITV, or FRTV neither Work, With or Without Turbo Lora,
Using RTX Upscaler or Not
@mrweaz also, i Resized my Image to 1024x1532 to Match the 32 Multiplier also didn't Work
Latest V 2.2
@RJY17 okay, try using the latest download, i just updated it. got a feeling you may have downloaded when the broken update was up.
@mrweaz thx, i'll test,
@mrweaz and now this RuntimeError: The size of tensor a (2688) must match the size of tensor b (21504) at non-singleton dimension 1
ive been trying to figure this out for 5 Hours, im tired, might as well get some sleep
@RJY17 that would mean you are using the wrong model, double check you are using the correct model for the correct type Ref/FLF. the workflow wokrs for me on a clean install. im using the int8 versions of the models hosted here
I have a problem: the final video I rendered is completely black—there is no picture or sound—even though the preview looked fine.
you need to update comfyui to the latest branch, not release v0.30.0 will not work. need the latest commit
The LightXTurbo (any lora actually) is giving me garbled audio i tried INT8_Convrot and the pruned version, what am I missing?
the new sampler sa_solver fixes bad audio
Error with the node: " Fearnworks Nodes " is not working 🥲
that node pack isn't even in the workflow, i swear my comfyui hates me, i don't have the node pack installed and everything is working for me. ill redo the whole WF, something has broken