Pass WAN_Model

Pass WAN_Model is a passthrough for the WANVIDEOMODEL type - the model object that Wan video workflows load and feed into sampling. One Wan model in, the identical one out, nothing else. If you build Wan video graphs, this is your junction node; if you don't, it's the easiest node in the pack to ignore.

Wan is the open video family that local generation consolidated on - Alibaba's 14B diffusion transformer plus a 3D causal VAE, frozen at Wan 2.2 and still the base a huge share of 2026 open video models run on. In ComfyUI, the Wan model comes out of its loader on a WANVIDEOMODEL-typed wire, and that wire is heavy: a 14B model is real VRAM, and you don't want multiple copies of it floating around your graph. A big Wan workflow tends to have one loaded model feeding several stages - a text-to-video pass, an image-to-video pass, maybe a refinement - and the standard failure mode is that the single model wire breaks when you bypass or swap the loader, taking every downstream branch dark with it.

That's the entire point of this node, and it's the same organizational trick as every other passer in the pack. The RvTools README frames passers as fixes for nodes that had problems getting information from bypassed inputs, and as the second node you need to form a managed group. Put Pass WAN_Model between your Wan loader and the fan-out, and you get one stable junction: swap or bypass the loader and you reroute a single wire instead of chasing half a dozen dangling connections. The mechanism is pure - model in, model out - no copying, no conversion, no reloading. The WANVIDEOMODEL type passes through untouched.

The schema

• model (WANVIDEOMODEL, required) - the Wan model object to pass through.
• Output model - the same object, unchanged.

No toggles, no widgets. The output name model matches the WANVIDEOMODEL inputs of Wan-compatible nodes, so it drops straight in.

Install

From ComfyUI-RvTools_v2. ComfyUI Manager → search "ComfyUI-RvTools_v2" → install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI-RvTools_v2

Under RvTools II / Passer. Note this node doesn't bring Wan with it - you still need the Wan video nodes and the (large, multi-GB) Wan weights separately. RvTools itself needs no model downloads, just standard pip deps (torch, numpy, Pillow, opencv-python, pilgram, pynvml, piexif).

Caveats

The pack is deprecated, superseded by ComfyUI_Eclipse - works, frozen. If you're not doing Wan video, skip this node entirely. And the pack-wide gotcha: saved workflows referencing RvTools v1 won't resolve because the original repo was pulled from GitHub; v2's rename is the fix.