Wan2.2 Resolution Presets 🎬✨

Wan 2.2 is picky about its canvas, and the failure mode is silent. Feed the 14B model a width that isn't divisible by 16, or the 5B model one that isn't divisible by 32, and you get a clip that's subtly wrong in a way that's easy to blame on the prompt - blocky edges, banding, the "why does this look smeared" effect. This node exists so you never hand-type those numbers again. It's three dropdowns that spit out the two integers your video workflow needs, pre-checked against the rules that actually matter.

The display name has a little more flourish than the job deserves - "Wan2.2 Resolution Presets 🎬✨" - but it's honest about what it's for. Wan 2.2 was the last open Wan (Alibaba took 2.5+ API-only), and in 2026 it's still the default quality-first local video base. This is a quality-of-life node for people who actually run it, not a headline feature. You reach for it the way you reach for a preset dropdown on an image model: because you want the officially-supported 1280x720, not whatever you guessed.

How it actually works

The interesting part lives in the frontend, not the Python. The node ships a curated master list of ~90 resolutions across 8 aspect ratios - 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 1:1, and 21:9 cinematic - and each entry is tagged rule16 or rule32. When you pick a mode, the bundled JavaScript filters that list in real time: the 14B mode shows only resolutions divisible by 16, the 5B mode only ones divisible by 32. The aspect-ratio dropdown re-filters from there, the resolution list sorts by pixel area, and it auto-picks a 720p/1280p option when it can.

The backend class is refreshingly dumb by comparison. get_resolution takes the "1280x720" string, splits it on the x, and returns (width, height). That's the whole compute. Every bit of intelligence is the curated list plus the filtering, which is exactly the right division of labor - and also why the frontend is a real dependency, not a nicety.

The inputs that matter

Only three, and only the first two really need your attention:

• mode - Wan2.2 - 14B Models (I2V/T2V) or Wan2.2 - 5B Model (TI2V). This one's load-bearing. Hover it and the tooltip tells you the two things everyone forgets: 14B needs the Wan 2.1 VAE, 5B needs the Wan 2.2 VAE. Different VAE, different divisibility rule. This dropdown silently changes everything downstream.
• aspect_ratio - eight options, dynamically filtered by mode. The 5B model has stricter constraints (divisible by 32), so it genuinely has fewer options; 21:9 is the thinnest list for both.
• resolution - shown as WIDTHxHEIGHT, sorted by total pixel area. It's the output you came for; you rarely need to touch it beyond the sensible default.

Outputs: width (INT) and height (INT). Wire them into the width/height inputs of whatever builds your latent - typically EmptyLatentVideo on a native ComfyUI Wan setup - and from there into your sampler. It also plays fine with image-preprocessing nodes that take width/height, per the README.

Installing it

No dependencies, no model files, no VAE bundled - pure dropdowns. Either search "Wan22 Resolution Presets" in ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/opj161/ComfyUI-Wan22ResolutionPresets.git

Restart ComfyUI and it appears under ImpactFrames💥🎞️/utils. The pack ships a second node, the legacy Resolution Selector, kept around so old workflows don't break - ignore it for new builds.

Where people get burned

• Wrong VAE. The node only outputs numbers; it can't load a VAE for you. If your 14B clip comes out washed out or grey, you grabbed the 2.1 diffusion weights and paired them with the 2.2 VAE (or none). The mode tooltip is literally there to tell you which to use - that's the whole game.
• Empty resolution dropdown. This is normal, not a bug. The list rebuilds when you switch mode or aspect ratio; give it a beat or flip the ratio.
• Expecting it to police you. It's a preset generator, not a validator. You can still wire a rule-breaking number into a latent manually; this node just makes it easy not to.