Kablex / ComfyUI-Ref2VA-VSA

—

OpenVDN / vdn-minimax-h3

VideoDeltaNet-H3: Live T2VA / I2VA / FL2VA / Ref2VA-like generation based on Minimax H3.

Zzz1phyrf / ComfyUI-H3-LongVideo

ComfyUI long-video controller for MiniMax H3 Ref2VA singing and speaking workflows

scottmudge / ComfyUI_MinimaxH3HybridLoader

A model hybridization loader to combine different layers/blocks of the fl2va and ref2va models

Saganaki22 / ComfyUI-Hyperflow

ComfyUI node pack for HyperFlow [Video Rebirth's] 8-step LoRA for MiniMax-H3 (video + audio), ported onto ComfyUI's native MiniMax-H3 model. No ComfyUI core files are modified; no custom sampler is needed.

AI-KSK / aiksk-minimax-h3-local-prompt

MiniMax H3 全域提示词工程 Skill v1.8.1｜AIMixer Director 导演台生产规则 + 武打/兵器/枪斗编排｜T2VA / I2VA / FL2VA / L2VA / Ref2VA｜Subject 与 Picture 语义分离、跨镜头因果桥、人物一致性、动作参考、音频控制、Prompt 修复｜ComfyUI / RunningHub / Codex

lololerigolo60 / Minimax-H3-prompt-studio

Local desktop app to build structured MiniMax H3 video prompts (T2VA/I2VA/FL2VA/L2VA/Ref2VA) using a local LLM via Ollama or LM Studio. Streams output, auto-describes reference images with vision models, 100% offline — no API keys, no cloud.

karuvanan / MiniMax-H3-Director-Cut-Studio

Premiere-inspired PySide6 director studio for MiniMax H3 Ref2VA with AI shot planning, semantic media enrichment, timeline prompt reconciliation and shot-aware long-video rendering through ComfyUI.

sepiablue-ai / ComfyUI-MiniMax-H3-W4A4-VSA

MiniMax H3 Ref2VA acceleration for ComfyUI on an RTX 4070 12GB: 4-bit FC1 weights + Streaming VSA sparse attention. 1024x1792 video with audio in about 3 min 17 s, about 3.4x faster than dense attention.

klfzqxs / ref2va-h3-video-optimizer

MiniMax-H3 Reference/Image-to-Video 提示词优化器 —— 用顺序迭代反馈方法把一段视频的提示词自动优化到更高质量版本，需本地 ComfyUI + 本地/云端LLM模型运行。

ANe5s / ComfyUI-MiniMax-H3-Hybrid

Enables the ComfyUI MiniMax H3 reference-to-video workflow to use a stronger FL2VA model and validated Ref2VA tensor-overlay modes to unlock MiniMax H3 capabilities.

sepiablue-ai / minimax_h3_workflows

ComfyUI workflows for MiniMax H3 (Ref2VA / FL2VA / VideoRef / Pose Control) at Full HD on 12GB VRAM.

xituisuany-max / dsh-client-ui-pet

DSH web GUI 鲸鱼娘桌宠插件：23 个序列帧动作、多吸附点、坐姿专属动作套、token 汇报、滑动选择器（RunningHub H3 Ref2VA 素材管线）

killkli / minimax-h3-lora-colab-skill

Installable Codex skill for training MiniMax H3 Ref2VA LoRAs on Google Colab