🪠️ WWAA Video Resolution

Video models are picky about dimensions in a way image models aren't. A resolution that works fine for one model will throw a hard error in another, because each architecture wants its spatial dimensions to land on a specific multiple - Wan and LTX want multiples of 32, Flux-class models are happy with 16, and so on. WWAA Video Resolution wraps all of that in dropdowns: pick your model, your resolution tier, aspect ratio, and orientation, and it hands you a width and height that are actually compatible instead of "should be fine."

This is the kind of node you don't appreciate until you've burned twenty minutes on a Latent size is not multiple of error mid-video-batch. It turns that failure mode into a configuration question.

How it works

The dropdowns:

• model_type - LTX2/Wan2.x, Q25xx, Flux.2, or Z-Image-Turbo. This sets the dimension multiplier: 32 for LTX2/Wan2.x, 112 for Q25xx, 16 for Flux.2, 64 for Z-Image-Turbo. This is the input that matters most - the wrong one is exactly how you get the "not multiple of" error.
• resolution - 480p through 4K tiers (with base heights of 512/704/1088/1408/2176, adjusted down to the model's multiple).
• aspect_ratio - 16:9, 21:9, 2:1, 4:3, 3:2, 4:5, 1:1. Square and portrait get special handling to keep proportions sane.
• orientation - Horizontal or Vertical (swaps width/height).
• use_custom + custom_width / custom_height - an escape hatch; flip use_custom on and the presets are ignored entirely and your exact integers pass through.

Outputs are width and height as integers, ready to wire into an empty latent, a video encoder, or a resolution-check node.

Install

Ships in WWAA-CustomNodes. ComfyUI Manager: search "WWAA Custom Nodes", install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/hgabha/WWAA-CustomNodes
# restart ComfyUI

No extra dependencies, no model files - it's pure arithmetic on the selected enum values, under the 🪠️ WWAA/Video menu.

Honest assessment

The multiplier table is the whole game, and it's a real convenience: remembering "Q25xx wants 112s" is not something you want to do from memory mid-workflow. But keep two things in mind. First, these multipliers are the author's tuned values, not a universal truth - always sanity-check the output against your actual model's documented constraints, especially as new model versions ship. Second, the presets are a starting point, not a law; the use_custom path exists precisely because real video work often wants a specific size no preset covers. For a workflow author, it's a great way to expose a clean "resolution" control on your graph that end users can't break - and for everyone else, it's the node that stops you from manually doing modular arithmetic on your video dimensions.