Get Float Extra

Get Float Extra is the float-typed read node in this pack's extras system: a Set Float Extra upstream stashes a decimal under a key in the image's image_info, and this node hands it back as a FLOAT you can wire straight into a CFG, denoise, LoRA-weight, or strength socket. If Get Int Extra is for whole numbers, this is the one for the values with the decimal points - which, in diffusion land, is most of the interesting ones.

The trick that makes it useful

The pack's image_info bundle carries your whole generation recipe, and its extras dictionary is the scratch space on top. Because image_info gets written into saved-image metadata and rebuilt by Image Reader, a value you set with Set Float Extra doesn't just live in the graph - it lives in the file. Load that image next week, and your CFG 7.5 or denoise 0.55 comes back out without anyone remembering to type it.

That's genuinely the right pattern for the tweak-values you tune a dozen times: set once at the top of the workflow, read wherever the sampler needs it, and let the metadata carry it into the next session instead of your notes app.

Using it

Inputs are the same shape as the rest of the family:

• extras - the IPT-ImageInfoExtras socket (from Image Info Context, a Set*Extra node, or another image-info node).
• key - the parameter-line key to look up, matching the writer exactly.

Output is float. Missing key or uncastable value gives you None rather than a crash, so it behaves like a dict lookup, not a promise. If your workflow can run without the value, wire a fallback; if it can't, the None will happily propagate until the sampler complains - which is its way of telling you the writer side isn't connected.

One caution

Watch the CFG/denoise ranges. The node doesn't validate; a value set with one workflow and read in another is just a float, and garbage floats (CFG 47, denoise 3.0) flow through unhindered. The troubleshooting knowledge base is full of "CFG too high / too low" stories, and this node is perfectly capable of being the delivery mechanism for one. Keep your writers and readers in the same pack, same ranges, and you'll be fine.

Installing it

Ships in the ComfyUI-Info-Prompt-Toolkit pack. Via ComfyUI Manager - search "Info Prompt Toolkit" - or:

cd ComfyUI/custom_nodes
git clone https://github.com/kinorax/comfyui-info-prompt-toolkit.git
cd comfyui-info-prompt-toolkit
pip install -r requirements.txt

Restart ComfyUI. No model downloads needed. Small single-author pack (GPL-3.0), modest community footprint - the extras pattern is its quiet superpower.