🔹 CYH Latent | Social Media Aspect Ratio

If you generate social content, you know the drill: Instagram square is 1080×1080, stories are 1080×1920, Twitter headers are a weird 1500×500, and remembering all of them on demand is a small recurring tax. This node is that list turned into a dropdown - twelve platform formats, each with the resolution baked into the label, producing an empty latent.

It's the social-media member of the Chye ComfyUI Toolset latent family, and the biggest preset list of the bunch. Everything else about it works like the pack's other aspect nodes.

How it works

Standard pack pattern: the label carries the size (Instagram Square (1:1) - 1080×1080), the node parses it, applies orientation and multiplier, rounds to a multiple of 32, and makes a zeroed tensor:

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

A LATENT out, straight into a KSampler. Nothing platform-aware about it - these are just resolutions, not exports; the node won't resize your final image to fit Instagram's upload rules.

The presets

• Instagram - Square 1080×1080, Portrait 1080×1350, Landscape 1080×608, Stories/Reels 1080×1920
• TikTok - Standard 1080×1920, Wide 1920×1080
• YouTube - Standard 1920×1080, Shorts 1080×1920
• Facebook - Feed 1200×630, Stories 1080×1920
• Twitter/X - Tweet 1200×675, Header 1500×500

That's a genuinely useful lookup table, and the 1500×500 Twitter header is the kind of number nobody keeps in their head.

The inputs that matter

• aspect_ratio - the twelve presets above.
• orientation - Portrait/Landscape, default Landscape. Same caveat as the rest of the pack: the default is static. The README again promises platform-smart orientation ("Portrait for Instagram Stories, Landscape for YouTube Standard") but the code just defaults the dropdown to Landscape - it won't flip when you pick a portrait format. Set it yourself, especially for anything 9:16.
• multiplier - 0.1–10.0. Watch the megapixels: most of these are right around 1–2MP. On a 1MP-native model, 1080×1920 (2MP) wants a ~0.7 multiplier and an upscale later; on newer 1–2MP-band models it's fine as-is.
• batch_size - 1–64. Genuinely nice here: build a batch of blank latents and run a whole platform's worth of images in one grid.

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. Deps: scipy, opencv-python plus requests/coloredlogs; torch and numpy are already there. Zip installs need the .git/.cnr-id file (Chye-ComfyUI-Toolset) per the README.

Honest verdict

The best use of this node is the batch workflow: pick a format, set batch to 4, get four images in the right aspect for your campaign in one pass. For a single image you could just type the numbers - but the header size you always forget is right there in the label.