WAN2.1 Resolution (Advanced)

The plain WAN2.1 Resolution node covers the four obvious cases: 480p and 720p, in landscape and portrait. The WAN2.1 Resolution (Advanced) node is the version for when "obvious" isn't enough - twelve presets including 4:3, 3:4, and square shapes, plus two intermediate sizes. If you're making vertical short-form content or anything that isn't plain 16:9, this is the one you want from the Image Size Tools pack (TheLustriVA/ComfyUI-Image-Size-Tools).

Why the extras matter

Wan 2.1 is a video model, and video models are allergic to two things: odd dimensions and memory surprises. Every one of the twelve presets is even on both sides, so your encoding pipeline stays happy. And while 16:9 is the default look, a lot of real use - phone-format vertical video, square social clips - wants a different box. The Advanced node is the pack's answer for those shapes, on top of the same 480p/720p family the base node gives you.

The mechanism, and the quirk you need to know

Same architecture as every node in this pack: a hardcoded dictionary in a single-file Python module, zero dependencies. Pick a label like "960×720 (4:3) - Classic 720p", get width, height, ratio, and halved dimensions back. Nothing is calculated at runtime except the halving.

Which brings us to the detail worth flagging: the halve_resolution checkbox does nothing. Read the source and halved_width/halved_height are computed unconditionally - the boolean is accepted as an input but never read. On or off, you get all five outputs. It's cosmetic in v0.3.4. The halved values are always there to grab, so the checkbox being dead costs you nothing - just don't rely on it to gate anything.

Inputs and outputs

• resolution - enum of twelve presets: the four core 16:9/9:16 sizes, 4:3 and 3:4 "Classic" variants at 480p and 720p, two 1:1 squares, and 1024×576 / 576×1024 intermediates.
• halve_resolution - boolean, decorative in the current release.

Outputs: width, height, aspect_ratio (STRING), halved_width, halved_height (INT). Feed width/height into your Wan video latent, or the halved pair when you're testing on 12GB-ish VRAM - 1280×720 on a 14B model is a commitment, and 640×360 lets you iterate. One note from the wider Wan community: 960×544, 1024×576, and 1280×720 are the shapes people actually report using for 480p/720p-class work, and the Advanced node's intermediates sit right in that band.

Installing

ComfyUI Manager → search "Image Size Tools", or:

cd ComfyUI/custom_nodes
git clone https://github.com/TheLustriVA/ComfyUI-Image-Size-Tools.git

Restart ComfyUI, no dependencies to install, no models to download. It's the same pack as the base WAN node - if you have one, you have both.

The honest take

If you're doing plain landscape video, the base node is enough and this is overkill. If your workflow touches vertical or square formats, the twelve presets make it the better default - one dropdown covers shapes you'd otherwise be hand-typing and hoping. The dead toggle is a small mark against polish, but since it never withholds the halved outputs, the practical verdict is: reach for this one for anything non-16:9, ignore the checkbox, and get back to generating.