Skip to content
Video

javawock7618/comfy-MiniMax-H3-workflows

Ready-to-use workflows for video and audio generation in ComfyUI

2 min readupdated

This repository provides a collection of optimized workflows for running the MiniMax-H3 model within ComfyUI (a graphical interface for creating generation pipelines). The set includes tools for first-and-last-frame video generation (FLF), reference-image video generation (R2V), video transitions (Bridge), face refinement (Face Refine), and text-to-speech (TTS). All workflows are adapted for INT8 quantization (compressing model weights to save memory), and support native attention optimizations alongside ControlNet integration (tools for pose and depth control). This release is ideal for videographers and enthusiasts looking to run generation locally with lower VRAM (video memory) requirements.

Key features

  • INT8 Inference: Using quantized model weights to lower video memory requirements and improve local execution efficiency.
  • Native Block Sparse Attention: Built-in attention processing optimizations in ComfyUI without needing extra custom nodes.
  • Spectrum and Lightx2v Turbo LoRA: Integrated sampling optimization and generation acceleration methods (LoRA is a small add-on to the base model).
  • Fun ControlNet Union: Support for controlling generation using OpenPose, Canny, and Depth maps in I2V and R2V pipelines.

What you can do

  • Create video by frames: Generate a clip based on the first and last frame (FLF/I2V) with optional upscaling and video continuation (Extend).
  • Transfer style from a reference: Run video generation using a reference image (R2V) with a specialized reference LoRA.
  • Connect two videos: Generate a smooth transition fragment between video clip A and video clip B using the Bridge workflow.
  • Generate speech: Use a separate workflow for audio synthesis (text-to-speech) based on MiniMax-H3 without launching the video generation process.

Example

To run it, you will need to install the base MiniMax-H3 model, download the required JSON workflow file for ComfyUI, connect the necessary custom nodes (such as for upscaling or face processing), and then load the INT8 format weights into the corresponding pipeline nodes.

Read moreDetails
• Includes workflows: FLF (I2V), R2V, Bridge, Face Refine, and TTS.
• Requires installing custom nodes: ComfyUI-Spectrum-MiniMax-H3, ComfyUI-H3-FaceRefine, Comfyui_Minimax_h3_latent_Upscaler, ComfyUI-Switchboard.
• Optionally supports Veda-on-ComfyUI for attention optimization in I2V and R2V.
• Designed to work with the MiniMax-H3 model from Comfy-Org.
Discuss 170

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # javawock7618/comfy-MiniMax-H3-workflows
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download javawock7618/comfy-MiniMax-H3-workflows --local-dir ~/models/comfy-MiniMax-H3-workflows
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • SageAttention 3v2.2.0from October 28, 2025

    even faster on RTX 50 / Blackwell

    FP4 attention for Blackwell. Faster than SageAttention 2, but the authors suggest v2 when accuracy matters.

    How to enable: Build from the sageattention3_blackwell folder of the repo; in ComfyUI via KJNodes.

    Earlier generations (2)
    • SageAttention 2++ — Quantised (INT8/FP8) attention with no visible quality loss. The main speed-up for image and video generation.
    • SageAttention 1 — First version: INT8 for Q·K. Still works, but 2++ is faster and more accurate.
  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (3)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
    • xFormers — The older way to speed up attention, works almost everywhere. Sage or Flash are faster where available.
  • cache-dit (DBCache)v1.5.6from November 20, 2026new release

    1.5–3× at nearly the same quality

    Reuses block outputs between diffusion steps. Supports most DiT models.

    How to enable: pip install cache-dit; one line cache_dit.enable_cache(pipe) for diffusers.

    Earlier generations (2)
    • MagCache — Skips steps by magnitude of change. More accurate than TeaCache, no calibration.
    • TeaCache — The first popular diffusion step cache. Works, newer methods are more accurate.
  • LightX2V step distill0.6.1from November 9, 2026new release

    5–10× fewer steps

    Distilled LoRAs and an engine: video in 4–8 steps instead of 30–50.

    How to enable: Load their step-distill LoRA, CFG 1.0, 4–8 steps.

  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

MiniMax H3 prompts

All prompts

Ready prompts with the resulting video: copy and try.

Model family

New posts about versions of comfy-MiniMax-H3-workflows (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

javawock7618/comfy-MiniMax-H3-workflows
Author
javawock7618
Category
Video
Released
Posted

On NeyroLearn

Hugging Face

0170

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments