🔹 CYH Latent | Phone Aspect Ratio

Everything on your phone is tall. Lock screens, wallpapers, Instagram's vertical grip on your attention - and generating images to fill those tall rectangles means setting up 1080×2340-style resolutions that feel wrong if you're coming from desktop 16:9 habits. This node hands you empty latents at real phone aspect ratios, portrait-first, from one dropdown.

It's the phone member of the Chye ComfyUI Toolset's latent family - same design as the Flux/SDXL/Qwen aspect nodes, but the presets are mobile screen sizes and the default orientation is Portrait because that's how phones actually get used.

How it works

Identical skeleton to the pack's other latent nodes: the dropdown label carries the resolution, the node parses it, applies orientation and multiplier, rounds to a multiple of 32, and emits a zeroed tensor:

latent = torch.zeros([batch_size, 4, final_height // 8, final_width // 8])

Output is a plain LATENT for your KSampler. No weights, no downloads.

The presets

• 16:9 (Standard) - 1080×1920
• 19.5:9 (Modern Tall) - 1080×2340
• 20:9 (Ultra Tall) - 1080×2400
• 18:9 (Tall) - 1080×2160

Those are real phone resolutions - 19.5:9 is the Pixel/iPhone class, 20:9 is the "ultra tall" crowd. The default orientation is Portrait, so for once the label and the output actually agree: 1080×1920 comes out as 1080 wide, 1920 tall.

The trap is the resolution, not the node

Here's the part the pack doesn't warn you about: 1080×1920 is 2 megapixels. On a model whose native band is 1MP (SDXL, or any 1MP-era model), feeding it a 2MP latent directly is how you get artifacts, not detail. The node will happily make the latent; the model may not.

For SDXL-era models, drop the multiplier to about 0.7 (→ 768×1360ish, right at 1MP) and upscale after if you need full phone res. For 2025+ models with 1MP–2MP native bands (Z-Image, Flux 2 Klein, Qwen-Image at ~1.3MP base), the full 1080×1920 is actually fine. Know what you're generating on, and pick the multiplier to match. The other two inputs are orientation (though you'll want Portrait for almost everything here) and batch_size (1–64) for wallpaper-set grid runs.

Install

ComfyUI Manager (search "Chye ComfyUI Toolset"), or:

cd ComfyUI/custom_nodes
git clone https://github.com/chyer/Chye-ComfyUI-Toolset
cd Chye-ComfyUI-Toolset
pip install -r requirements.txt

Restart ComfyUI. The notable dependencies are scipy and opencv-python; torch and numpy are already yours. If you installed from a zip, add .git/.cnr-id containing Chye-ComfyUI-Toolset per the README to avoid workflow-load errors.

Honest verdict

Convenient if you batch phone wallpapers; the multiplier guidance is the part to remember. It's a preset picker, not magic - but a preset picker for tall resolutions is a thing you'll use more than you'd think once it's in your node list.