AIAny
AI Video2025
Icon for item

ComfyUI-WanVideoWrapper

Wraps Alibaba's Wan text/image-to-video models and many derivatives (SkyReels, WanAnimate, MultiTalk, Phantom) as ComfyUI nodes, adding FP8 scaling, block swapping and context windows for 1000+ frame clips.

Introduction

Video-generation research now ships faster than any single app can integrate it, and that gap is exactly what this project fills. It's the staging ground where new Wan-family models and one-off research techniques get usable ComfyUI nodes weeks before native support exists — the maintainer openly calls it a personal "sandbox" and tells you to switch to native ComfyUI nodes once they catch up.

What Sets It Apart
  • Breadth over polish: it tracks an unusually wide swath of the Wan ecosystem — base T2V/I2V plus SkyReels, ReCamMaster, Phantom, WanAnimate, VACE, MultiTalk and HuMo — so a new paper's weights often have a working node here first.
  • Memory engineering, not just plumbing: FP8 scaling, block swapping, and torch.compile support let big models run on consumer GPUs, and context windows push clips past 1000 frames instead of the usual few seconds — one tested config rendered 1025 frames under 5GB of VRAM.
  • Honest about its role: rather than competing with ComfyUI core, it explicitly defers to native implementations once they stabilize, keeping its focus on what isn't yet supported.
Who It's For

Great fit if you chase the latest Wan and video-diffusion releases, like building custom node graphs, and accept some breakage in exchange for early access. Look elsewhere if you want a stable, batteries-included video tool — for mature models the native ComfyUI nodes are simpler and better maintained, and the sheer number of supported models here means workflows can shift under you between updates.

Information

  • Websitegithub.com
  • AuthorsKijai
  • Published date2025/02/25

Categories

More Items

Hugging Face
AI Video2026

Provides Parallel Decoding Distillation (PDD) LoRA adapters that accelerate MiniMax-H3 video generation into few inference steps. Includes official 8-step Acc LoRAs for FL2VA and Ref2VA (rank=64, network_alpha=64, BF16), demo comparison videos, and example scripts using Diffusers' MiniMax-H3 ModularPipeline.

Hugging Face
AI Video2026

Conditions a MiniMax‑H3 video generator with a single ControlNet‑Union checkpoint to accept Canny, Depth, HED, MLSD or Pose control videos and run video inpainting. Guidance‑distilled for one‑pass inference; requires the base MiniMax‑H3 weights and specific control-branch config.

Hugging Face
AI Video2026

Upscales Minimax H3 24-channel VAE latents in-place to increase spatial resolution while preserving the time dimension. Replaces the decode→pixel-upscale→encode round-trip with a learned 2D/3D latent upscaler to save compute and avoid interpolation ghosting; supports 1.0–4.0× scaling.