S1Z35 - Hunyuan

The Hunyuan node in this pack covers Hunyuan Image, Tencent's image model - and if you're here, you probably already know why a resolution dropdown for it is a genuinely useful thing. Hunyuan Image is the heavy one in the room: an 80B image MoE that wants an unreasonable amount of VRAM, the kind of model where every byte of latent you shave off matters. This node gives you its native-ish sizes in a dropdown, and it also ships with a typo in its preset file that will hand you a nonsense resolution if you pick the wrong entry. Let's fix that first.

The bug to know about

presets/hunyuan.csv ships with this line:

250, 1440, 16:9

That's almost certainly meant to be 2560, 1440. Select the "16:9" preset and you get a 250×1440 latent - a bizarre near-strip that no model wants and that will waste a generation before you notice why. Edit the file, fix the line, then delete the node and re-add it (or restart) so the dropdown rebuilds. Consider it a rite of passage for this pack.

What it is

A size-preset node from the ComfyUI-S1Z35 pack for Hunyuan Image. The hunyuan.csv ships with six entries: 2048×2048 square, the broken 16:9, a 1.6667 bucket at 2560×1536, plus 3:2, 4:3, and a 3:2 HD at 2400×1600. These run big - the square alone is 2048 - which is the point: a model this heavy is exactly where you want a dropdown telling you the sizes it was built around instead of free-typing numbers.

Honest context: Tencent's Hunyuan program is the one that keeps shipping frontier-sized open weights nobody can comfortably run - the 80B image MoE needs a server, not a gaming card. So this node serves a niche of a niche, and you're not missing much if it's not your daily driver.

How it works

Same shared engine as the rest of the pack. At node creation it reads presets/hunyuan.csv, parses each width, height, label line into a dropdown entry, and on selection extracts the numbers back out. No model loading, no dependencies, no GPU work.

Inputs and outputs that matter

• preset - the dropdown of Hunyuan sizes. Set it, done.
• swap_dimensions - boolean, off by default. Flips width and height for instant portrait versions of any bucket.

Outputs are width (INT) and height (INT), wired as usual into EmptyLatentImage or any size input ahead of the sampler.

Installing it

Same light install as the rest of the pack - no requirements.txt, no model files to download:

cd ComfyUI/custom_nodes
git clone https://github.com/K3nt3L/ComfyUI-S1Z35

Restart ComfyUI and it appears under S1Z35/Resolution. Or use ComfyUI Manager → Install Custom Nodes and search for "S1Z35".

Common issues

• The 16:9 typo - see above. Fix it, or don't select that entry.
• Edits don't show up until you refresh the node. The dropdown builds at node creation; edit the CSV, delete and re-add the node (or restart).
• "Missing hunyuan.csv (512x512)" - the fallback when the file is gone or empty.
• OOM at 2048²? That's the model being a monster, not the node. If Hunyuan Image won't fit your card, no preset node will save you - you're at the hardware limit, and the dropdown can't make an 80B MoE lighter.

It's a small, single-purpose node with one obvious defect in its shipped data. Fix the typo, and it does exactly what it says: Hunyuan Image sizes at a glance.