No Preview Save Image
Save every frame without drowning in previews
- images
The name is slightly backwards: this node saves everything, it just refuses to show you. No Preview Save Image is ComfyUI's built-in SaveImage with the preview stripped out - it writes PNGs to your output folder exactly like the core node, but returns no UI result, so the canvas doesn't fill up with thumbnails while a batch runs. If you've ever queued 500 generations and watched ComfyUI grind to a halt rendering previews of every single one, this is the node you were looking for.
How it works
Read the source and you'll see it's essentially core SaveImage's save_images with the preview-facing bits cut away: it returns an empty {"ui": {}} dict, which ComfyUI reads as "nothing to show." Everything else is untouched. It writes to the standard output directory, uses the same {filename_prefix}_{counter:05}_.png naming, compresses at level 4, and - the part people forget - embeds the full prompt and workflow as PNG metadata by default, so drag-and-drop workflow reconstruction still works after the fact.
That makes it a drop-in replacement you can swap in at batch time without changing your naming or your folder layout. If a core SaveImage already named a file ComfyUI_00001_.png, this node picks up the same counter.
The inputs
images- the IMAGE tensor(s) to save. It'll happily save every frame in a batch.filename_prefix- prefix for the files, defaultComfyUI. Give it something meaningful when you're batch-running, or you'll be sorting through hundreds of same-named folders later.save_metadata- default true. Leave it on. Turning it off strips the workflow from the PNG, which breaks the "drag it back into ComfyUI" trick and makes reproductions much harder to reconstruct.
It has no outputs - it's an output node, the end of the line. It also quietly accepts ComfyUI's hidden prompt and extra_pnginfo inputs, which is how the metadata gets embedded without you wiring anything.
Installing it
Same pack, same trivial install: qbug-batch has no dependencies, no models, no API keys - just standard library plus ComfyUI's built-ins, which is exactly the kind of pack you want running arbitrary code on import. Install via ComfyUI Manager (search "qbug-batch") or:
cd ComfyUI/custom_nodes
git clone https://github.com/Q-Bug4/comfyui-qbug-batch
Restart and it lives in the "image" category.
The tradeoff
The whole point is that you get no visual feedback, and that's both the feature and the hazard. The first time you run it you will absolutely wonder if anything is being written - check ComfyUI/output/. My habit: keep a development copy of the workflow with the normal SaveImage wired in, get it working and tuned, then swap in this node and launch the batch. You lose the per-image glance, but you gain a queue that doesn't stutter and an output folder you can sort and grade afterwards. For batch work, that's usually the right trade.
One more thing: because there's no preview, nothing shows in the browser's history panel either. If you rely on that to remember what ran, keep a text log or a good filename prefix. It's a small cost for not having ComfyUI choke on your cross-product batch.
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| images | IMAGE | — | |
| filename_prefix | STRING | ComfyUI | — |
| save_metadata | BOOLEAN | true | — |
Outputs (0)
No outputs