Video catalogue · Verified local path

MiniMax-H3-Fun-Controlnet-Union local guide

ControlNet-Union branch for MiniMax H3 that guides local video generation from Canny, depth, HED, MLSD, pose or inpaint control videos.

Choose an app

Start with Recommended. No terminal commands are shown.

Compare video models
Open model filesRecommended verified starting pointOpen in UnslothLaunches Unsloth Desktop on this modelGet ComfyUI DesktopOfficial desktop app for this compatible modelOpen advanced setupPython runtime for experienced usersCheck my machineSee whether this model fits your saved hardware

Desktop app links require the app to be installed. If nothing opens, LocalClaw will show app-download and model-file fallbacks.

What it does

Video To VideoControlled Video GenerationText To VideoInpainting

The official model card publishes a 6.8 GB control-branch safetensors checkpoint and VideoX-Fun scripts for Canny, depth, HED, MLSD, pose and inpaint workflows. It loads on top of MiniMax H3; the authors note the transformer plus Qwen3-VL text encoder do not fit fully on one 80 GB GPU, so LocalClaw records 80 GB NVIDIA VRAM with offload and 128 GB system RAM as the conservative workstation floor.

Hardware figures are practical entry floors, not performance guarantees. Resolution, duration, precision, offloading and runtime versions can materially change memory use.

Strengths

  • One control branch covers five control-video types plus inpainting
  • Official Alibaba PAI checkpoint and VideoX-Fun code path
  • Guidance-distilled workflow runs at guidance scale 1.0

Limits to know

  • Requires separate MiniMax H3 base weights
  • Community license has territorial and acceptable-use restrictions
  • Workstation-class NVIDIA memory is required for practical local use