🪠️ WWAA JPEG Preview
See your image's real JPEG file size before you commit to saving it
- image
Every image you generate looks great in the preview pane. The number you actually care about - how many kilobytes it'll be as a JPEG at a given quality - is the one ComfyUI won't tell you. WWAA JPEG Preview fixes that specific gap: it converts your image to JPEG in memory, shows the preview, and stamps the dimensions and file size in KB underneath.
It's an output/display node, so it has no outputs to wire - you drop it at the end of a branch like any preview, and the info appears in the node's own UI. If you're generating for the web, for sharing on Discord, or for any pipeline where file size matters more than "does it look nice at 100% zoom," this is the difference between guessing at quality settings and knowing.
How it works
Under the hood it's simple: the image tensor gets converted to a PIL image, encoded as JPEG in a memory buffer at your chosen quality, and the resulting byte count becomes the KB figure. It also writes a temporary JPEG out to ComfyUI's temp directory so the UI has something to display - that's what the preview you see actually is, an already-compressed version rather than the raw output. It's worth internalizing that the preview you're looking at is the JPEG; if it looks worse than your raw image, that's the compression talking, and that's real information about what your export will look like.
The one input that matters is quality (1–100, default 70, steps of 5). Higher is better quality and larger files; 70 is the classic web-photo default, and you can drag it and watch the KB count move in real time to find the quality-to-size sweet spot for whatever you're posting. That's the whole workflow: dial quality, read the size, done.
Install
It's part of the WWAA-CustomNodes pack. Easiest path is ComfyUI Manager - search "WWAA Custom Nodes" and install. Manual route:
cd ComfyUI/custom_nodes
git clone https://github.com/hgabha/WWAA-CustomNodes
# then restart ComfyUI
No extra Python dependencies (the pack's requirements.txt says so explicitly - numpy, Pillow, and torch, all already there), no model files to download, nothing to configure. It shows up under the 🪠️ WWAA menu.
Notes and gotchas
This is a niche tool, and it knows it. If you mostly generate and keep PNGs, you'll barely need it - but the moment you start exporting web-friendly JPEGs, having the file size at a glance beats saving, checking in Explorer, deleting, and re-queuing. One thing to keep in mind: the node encodes only the first image of a batch, so if you're feeding it 12 frames to check export sizes, you're only seeing frame 0's numbers. For single-image quality tuning it's perfect; for batch export sizing you'd want the sibling Save JPEG node instead. It's from the small WeirdWonderfulAI.Art team, and it's one of those "written because we hit this exact annoyance" utilities - low ambition, real convenience.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | — | |
| quality | INT | 701–100 | — |
Outputs (0)
No outputs