Quick Save Image 💾 (TensorVizion)

A ComfyUI PNG already carries its workflow - the graph, the prompt, the seed, all embedded as text chunks in the file. That's why dragging an old output back onto the canvas rebuilds the whole thing. Quick Save Image takes that convention and adds the pieces core SaveImage doesn't give you: a dedicated seed, a model name, and a notes field, stamped into the same PNG. You don't just save the picture, you save what made it.

What it does

It delegates the actual save to ComfyUI's core SaveImage, so your image lands in ComfyUI's managed output tree with the standard prefix_00001_ naming and shows up in the gallery immediately. Then it re-opens that file and appends three PNG text chunks:

• tv_seed - whatever seed you wired in (useful when your graph's real seed lives elsewhere than the workflow JSON).
• tv_model - the model name you typed in.
• tv_notes - your freeform notes, if you filled them in.

Crucially, it preserves the existing workflow and prompt chunks rather than replacing them - so the "drag it back onto the canvas" property survives intact. The metadata enrichment is best-effort: if the re-write fails for any reason, the core save has already succeeded, so you still get your image, just without the extras.

The inputs

• images - the IMAGE(s) to save. First frame of the batch is what gets written.
• filename_prefix - default "ComfyUI". Supports formatting like %date:yyyy-MM-dd% and can pull values from other nodes, e.g. %Empty Latent Image.width%.
• seed - the INT to embed.
• model_name - a STRING, whatever label you want recorded.
• notes - optional multiline STRING.

Output is a single saved_path STRING with the full path of the file it wrote - genuinely handy if you're building a workflow that needs to know where its output went.

Install and gotchas

Same pack, same install:

cd ComfyUI/custom_nodes/
git clone https://github.com/TensorVizion/OmniNodes

Restart ComfyUI, or search "OmniNodes" in ComfyUI Manager. No extra dependencies - it's a thin wrapper over core SaveImage plus Pillow, which ComfyUI already ships.

The honest caveats: saved_path is only populated if ComfyUI can resolve the output path from its UI metadata - usually yes for normal desktop saves, but don't build a critical pipeline on it. And the chunks it writes use tv_-prefixed keys, so they're readable by this pack's Metadata Reader, not by A1111-style tools that look for a parameters chunk - if you need CivitAI's uploader to auto-read your settings, that's a different metadata dialect than what this node writes. Also note the seed it stamps is whatever you feed it; it doesn't magically read the KSampler's seed, so wire the same seed node to both and they'll agree.