DepriveD - MiniMax H3 All-in-One HIGH-SPEED Workflow
A simple, fast, high-quality MiniMax H3 workflow designed for consumer cards.
This is a dual-mode All-in-One:
T2V / FFLF path
Reference path (images + videos + audio)
A single switch at the top forces only one path to be active at a time so you can’t accidentally run both.
Core Features
7 + 3 two-pass latent upscale (using minimax_h3_latent_upscaler_3d_bf16.safetensors)
Total of only 10 steps for high-resolution results that look like they were generated fully at the final resolution
You get a quick first-pass preview you can examine before the second pass finishes — so you can see if the video is going in the direction you want without waiting for the full generation
Optional FPS boost (leave at 24 to disable, or set to 30/48 etc.)
Optional RTX Video Super Resolution on top (2.0× recommended, 1.5× also works well)
Clean parameter panels — everything important is exposed, the rest is hidden
Preset resolutions (all correctly divisible by 32 for the latent upscaler)
Simple duration slider
Optional LoRAs
Full media loader for the Reference path (9 images / 3 videos / 3 audios) with resize, crop, trim controls inside the panel
How to Use (Quick Start)
Choose Resolution (or click the wheel to set your own — must stay divisible by 32)
Choose Duration
(Optional) Add LoRAs
Write your prompt
Hit Run
That’s it. Everything else has good defaults.
Recommended Starting Settings
First Pass Steps: 7 (6 or 8 looks good too)
Second Pass Steps: 3 (you can try 2 for a bit more speed)
First Pass Denoise: 1.0
Second Pass Denoise: 0.2–0.3
Turbo LoRA Strength: 1.0 (lowering requires more steps)
Video Shift: 12.0
Audio Shift: 3.0
Increasing steps much higher with the Turbo LoRA will start burning the image. The starting values above are a strong baseline.
Important Notes
Pictures do not save automatically. Right click any of the video outputs to save, OR set them to save instead of preview.
The entire complex wiring is hidden behind the LoRA panels on each path.
This is a new model and the community is still experimenting — feel free to try different settings.
Between Sage Attention, the latent upscaler, and the Turbo LoRA, generation speed on a 5070-class card is competitive with (and in many cases faster than) typical LTX workflows while keeping strong quality.
Required Custom Nodes
Pixaroma
ComfyUI-VideoHelperSuite (VHS)
ComfyUI-Fantastic-MiniMaxH3-PromptBuilder (Fantastic H3 Media Loader)
VRGamedevgirl (Required for latent splitting and seperating.)
Reroute nodes (usually included with base ComfyUI or common utility packs)
Required Models
T2V-FFLF: minimax_h3_fl2va_pruned_int8_convrot
Reference: minimax_h3_ref2va_pruned_int8_convrot
CLIP: qwen3vl_32b_minimax_h3_nvfp4_awq (or your preferred quant)
Video VAE: minimax_h3_video_vae_fp16
Audio VAE: minimax_h3_audio_vae_fp32
Turbo LoRA: minimax_h3_turbo_...
Latent Upscaler (required): minimax_h3_latent_upscaler_3d_bf16.safetensors
Enjoy fast, high-quality MiniMax H3 generations!!!!
P.S.
PLEASE let me know anything I can do to IMPROVE or FIX the workflow, I'll update as fast as possible!
Description
FAQ
Comments (23)
also LBH-123-AI/Minimax_h3_latent_Upscaler needed in addition to VRGamedevgirl nodes, not pulled automatically.
@Emanresu_ym correct, very good. some peope recently reported an issue with it NOT donwloing ALL of the nodes if you use comfyui manager. the WAY to fix that untill it's fixed on the manager's end is to use git clone throught a cmd prompt in yout custom nodes folder BUT, you need to COMPLEELY delete the VRgamergirls folder first and start from scratch. If you explain this to Grok, Chapt GPT, or Google AI, they can have it fixed for you in five minutes i'm sure, the comfyui manager is NOT installing the nodes properly with the current version of the manager. Keep digging, and let me know.
@DepriveD_AI for me it is all working after those 2 git pulls
@Emanresu_ym That's my man!!!! I'll be updating soon and attempting to mak it cleaner and more stable.
ive been trying to fix an error with the comfyui-vrgamedevgirl nodes, cannot get it to start cause
VRGDG_MiniMaxH3LatentUpscaleModelLoader
VRGDG_MiniMaxH3LearnedLatentUpscale
VRGDG_MiniMaxH3ReplaceUpscaledVideoLatent
all are apparently missing even though all nodes are loading normally. spent over an hour, no progress.
yes. those are part of th VRgamergirl nodes (I think that's what it's called). did you use the manager to install custom missing nodes? if THAT doesn't work, uninstall the ENTIRE thing and do it from scratch with a git clone. I PROMISE google AI, chat GPT, or Grok can walk you through that and in can be fixed and working in 5 minutes. Just explain the situation to your AI of choice if you cant't find it or fix it via the comfyui manager.
Also have you snuck behind the lora panel to find the hidden subgraph with the wiring where these nodes are located? go back in there and see if there's any red boxes. let me know! I'll help any way i can.
@DepriveD_AI dawg ive been chatting with grok on 3 accounts. ive tried git cloning and checking all the dependencies, installed directly via comfy, nothing seems to work.
ill try again tomorrow.
have you looked in your cusotm nodes folder amd found the vrgamedevgirl folder? if ya did, inside the folder see if the node names you are missing are there. IF THEY ARE NOT THERE. DELETE the folder completely, than ask your AI for the cmd prompt download. the newest version. make sure the cmd prompt is from your custom nodes folder. A complete fresh reinstall just worked for some others who got at me personally.
Did you update comfyui?
okay its fixed, a fresh reinstall with git clone fixed everything. this was tried before, but for some reason now its working perfectly. thank you for the workflow, very fast and great quality.
@104point That's what I like to hear guy!
Hi! First of all I want to thank you for your workflow. It's fast and provides good outputs.
However, I keep trying different resolutions and at some of the "standard" ones (1280*720) the second pass (for I2V-First frame only) gives this error:
RuntimeError: shape '[1, 24, 1, 1, 22, 2, 40, 2]' is invalid for input of size 86400
First pass also gives similar error in other resolutions for example 4:3 (960*720).
Do you know how to fix this?
hmmm... ok FIRST things first. any idea what the resolution of your beginning image is? if you CLICK in the settings wheel of the image in th workflow you can ses the RESIZE OPTION for the image. set the long side to 1280. let's start there! Let me know if that helps, if not I'll do some more diggin' for ya. The LATENT is VERY sensative with the (divisible by 32) rule. The resolution is getting divided by two for the first pass, IF THOSE 2 numers aren't divisble by 32 it will fail when it starts the second pass.
@DepriveD_AI Thank you! It's working like this. I guess my bad that I didn't check the Resize option. I just uploaded the image and chose resolution, without resizing.
Damn, reaallllyyy nice wf m8
Thank ya my guy!
Can't get vrgamedevgirl nodes to work. The repos are 404. managed to install through comfyuimanager yet they still give errors
Missing Node Packs
Install missing packs to use this workflow.
To install missing nodes, first run pip install -U --pre comfyui-manager in your Python environment to install Node Manager, then restart ComfyUI with the --enable-manager flag.
VRGameDevGirl Custom Nodes3
VRGDG_MiniMaxH3LatentUpscaleModelLoader
VRGDG_MiniMaxH3LearnedLatentUpscale
VRGDG_MiniMaxH3ReplaceUpscaledVideoLatent
1Missing node type
A node type is missing or unavailable. The workflow may be corrupted or require a custom node.
I had to delete the comfui-vrgamedevgirl folder and redownload it. There is a file you are probably missing called VRGDG_MiniMaxH3LatentUpscaler.py
@grokjunkie161 it was probably that, spent 2 hours with 2 ai's treoubleshooting this :(
Grok is right. For SOME reason the current version of comfyui manager OR the nodes themselves ARE NOT updating properly. Hopefully that changes soon. For now COMPLETELY deleting the folder from your custom nodes AND a fresh reinstall should work.