Get Size Extra
Pull a resolution out of image metadata, not your memory
- extras
- width x height
- width
- height
Get Size Extra reads a stored resolution out of an image's image_info extras and hands it back as proper size sockets. A Set Size Extra upstream writes a width/height pair under a key; this node unpacks it into a combined width x height output plus separate width and height integers. It's the size-flavored member of the pack's getter family, and the one that quietly saves you from hardcoding resolutions in a dozen places.
Where sizes live in this pack
This pack is built around reusable image_info bundles, and resolutions travel inside them - which matters because ComfyUI workflows have a way of growing a second, third, and fourth resolution knob. The pack's Aspect Ratio to Size node resolves a size from a ratio and an anchor; the trim variant adds margins; and the Set/Get size extras let you park that resolved size in metadata and reuse it wherever a size is needed. Generate at 1152×896, stash it, and the same value flows into the sampler, the upscaler, and the next workflow without you reconciling numbers by eye.
Inputs and outputs
Two inputs, same shape as every getter in the family:
extras- theIPT-ImageInfoExtrassocket (fromImage Info Context, aSet*Extra, or another image-info node).key- the parameter-line key to look up.
Outputs are three:
width x height- the combinedIPT-Sizesocket, which slots straight into size-accepting nodes (including the pack's own).width- theINT.height- theINT.
The combined output is the one you'll use most; the split ints are there for when a node only wants one dimension or you need to do math. If the key's missing, you get None on the ints - handle it downstream or make sure the writer side is connected.
Resolution sanity, briefly
A stored size is only as good as the size you stored. The troubleshooting side of the ecosystem has one loud, repeated lesson here: generating at resolutions far outside a model's training range invites tiling and duplicated anatomy. Get Size Extra will faithfully reproduce whatever you gave it, including the 2048×2048 mistake. Store sizes that make sense for the model you run, and let the node be the boring messenger it's meant to be.
Installing it
Part of the ComfyUI-Info-Prompt-Toolkit pack. ComfyUI Manager - search "Info Prompt Toolkit" - or:
cd ComfyUI/custom_nodes
git clone https://github.com/kinorax/comfyui-info-prompt-toolkit.git
cd comfyui-info-prompt-toolkit
pip install -r requirements.txt
Restart ComfyUI. No models to download. Small single-author pack (GPL-3.0) with a modest community footprint - the extras pattern is its strength, and this node is that pattern aimed squarely at resolutions.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| key | STRING | Parameter line key | |
| extrasopt | IPT-ImageInfoExtras | — |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| width x height | IPT-Size | — |
| width | INT | — |
| height | INT | — |