NAI Meta Apply

If you've been generating in NovelAI's web client for a while, you almost certainly have a folder of PNGs whose settings you'd love to have back without retyping them. NAI Meta Apply is the "reapply the recipe" button: drop a NovelAI image onto it, pick which parts of its metadata you want, and those values get pushed into your current workflow. It's the reuse half of the ComfyUI-4A-NovelAI pack's metadata story, and it's the node you reach for when "I made this good thing once and want to make more like it" turns into "wait, what was the CFG on that one?"

How it works

This is a UI-only node - it has no outputs, and it doesn't call NovelAI's API at all. Under the hood the pack's backend parses the metadata embedded in the PNG (the tEXt/iTXt chunks every NAI image carries, same family of chunks ComfyUI uses for its own workflows - see the KB's image-io-metadata doc) and, depending on which checkboxes are on, sends those values to the Prompt Scheduler and Sampler. Metadata reading is local and free: no token needed, no Anlas spent. Only generating a new image costs anything.

The inputs that matter

The whole node is a toggle board. The ones a beginner actually touches:

• Prompt - apply the image's positive prompt (on by default).
• UC - apply the negative. "UC" is NovelAI's name for the negative/unconditional prompt, so don't go hunting for a "Negative" checkbox.
• Characters - apply the character set (off by default).
• └ Append - instead of replacing your current characters, append the image's to them. Only meaningful with Characters on.
• Settings - model, sampler, steps, CFG and friends (on by default).
• Seed - apply the original seed (off by default, which is usually what you want when you're iterating).

There's also a hidden image string field - that's the drag-and-drop slot, not something you type in.

Where people get burned

The README is careful about two boundaries, and both are easy to hit by accident. First, applying Settings deliberately does not overwrite resolution - the pack doesn't want a 1216×832 generation silently forcing itself on your current canvas. Second, applying only the positive prompt preserves your Scheduler's character, action, and scene fields, which is great when you only want the wording - and a surprise if you expected the whole prompt to be replaced.

The other classic trap has nothing to do with the node: re-encoded PNGs carry no metadata. If your "NovelAI image" came from a Reddit post or a Discord preview, the chunks are long gone and this node has nothing to read. Non-NAI images can still be inspected for positive/negative text, but their model and generation settings aren't interpreted as NovelAI settings. Keep the original PNGs - the copy that came out of the generator is the only one that round-trips.

Install

Same pack install as the rest of ComfyUI-4A-NovelAI: ComfyUI Manager (search 4A NovelAI), or

cd ComfyUI/custom_nodes
git clone https://github.com/tsukino4a/ComfyUI-4A-NovelAI.git ComfyUI-4A-NovelAI
cd ComfyUI-4A-NovelAI
python install.py

then restart ComfyUI. Only real dependency is msgpack; everything else comes from ComfyUI. Once installed, drop this node next to your Scheduler and Sampler and start pulling old generations back to life.