GR Image Details Saver

This is a sink node - you plug your final images into it, give it a filename prefix, and it writes them to disk. That's the whole job, and the node has no outputs because there's nowhere left for the data to go. If you've used ComfyUI's own Save Image node, the shape will feel familiar: this is GraftingRayman's take on the same idea, just under his own naming so it slots into his other pack nodes without you reaching outside it.

How it works

Wire an IMAGE in, set a filename_prefix, and every time the graph runs, the node writes the batch to your ComfyUI output folder using that prefix (default GR_) plus whatever numbering scheme ComfyUI's save path uses to avoid collisions. Because it's marked as an output node internally, ComfyUI treats it as a terminal point in the graph - it'll always execute if it's reachable from your queue, even if nothing downstream depends on it, which is exactly what you want from a save step.

The "Details" in the name is the tell: nodes in this family typically piggyback the generation's metadata - prompt, seed, sampler settings, whatever the workflow carries - into the saved file the same way ComfyUI's default save node embeds workflow JSON in the PNG. That's what makes an image "draggable" back into ComfyUI later to recover the exact graph that made it. Treat that as the expected behavior rather than a guaranteed spec, since the pack doesn't document the internals - but it's standard practice for this class of node, and it's the whole reason a "details saver" exists instead of a plain save.

The inputs that matter

• images (IMAGE, required) - the batch you want written to disk. This is the only thing that actually needs a wire; everything downstream of your sampler and VAE decode funnels in here.
• filename_prefix (STRING, default GR_) - what your output files get named. Change it per-project so you're not hunting through a folder of identically-numbered GR_00001.png files six months from now. A prefix like character-sheet_ or client-batch3_ earns its keep the first time you need to find something fast.

No outputs - this is where the graph ends.

How to install it

Easiest path is ComfyUI Manager: search "GraftingRayman" and install the pack, then restart ComfyUI. Manually, drop into your custom nodes folder:

cd ComfyUI/custom_nodes
git clone https://github.com/GraftingRayman/ComfyUI_GraftingRayman

then restart. There's one dependency the author calls out for the pack as a whole, and it's not optional: you need OpenAI's CLIP package installed separately from whatever CLIP ComfyUI already ships with.

# portable build
.\python_embeded\python.exe -m pip install git+https://github.com/openai/CLIP.git

# system python
pip install git+https://github.com/openai/CLIP.git

This isn't specific to the Details Saver - it's a pack-wide import dependency. Skip it and none of the GraftingRayman nodes will show up in your node list, not just the ones that would obviously need CLIP.

Common issues & troubleshooting

None of the GraftingRayman nodes appear after install. This is almost always the missing CLIP package, not a broken git clone. The pack's __init__.py imports it up front, so if that import fails, the whole node registration silently fails with it. Run the pip command above for whichever Python your ComfyUI install actually uses (portable vs system), then restart.

Files aren't where you expect. Double-check you're looking in ComfyUI's configured output directory, not the custom node's own folder - save nodes in this pack write to the standard ComfyUI output path, keyed by your filename_prefix, the same as the built-in Save Image node.

Nothing happens when you queue. Make sure images is actually wired from something upstream (a VAE Decode, typically) - an unconnected required input will stop the node, not silently skip it.