Image Info

If you've ever looked at an A1111 screenshot with its neat little info block - seed, steps, CFG, sampler, the whole recipe - and wanted that inside a ComfyUI graph, this is the node. Image Info takes your checkpoint name, prompts, resolution, sampler settings, LoRAs, and VAE, and formats them into a single readable string. Feed that into a text display node or save it alongside your image, and you've got the "here's exactly how this was made" caption without leaving the graph. The README credits WLSH's nodes as the inspiration, which tells you the lineage: this is a metadata formatting tool, not a magic anything.

How it works

There are two halves. Half of the inputs arrive by wire - checkpoint_name, posetive (yes, that typo is the real input name, it's misspelled in the code), negative, width, height, plus optional vae_name, lora_list, manual_input, and extra_info_input. The node stitches those into a block that looks like:

Checkpoint name: ...
VAE Name: ...
Loras: Lora1: my_style, Strenght: 0.7, Strenght Clip: 0.7
Seed: ..., Resolution: 1024x1024, Steps: ...
CFG: ..., Sampler: ..., Scheduler: ..., Denoising: ...
Posetive Prompts: ...
Negative Prompts: ...

The other half is where it gets clever and then fragile. The required info input is typed INFO, and the code reads keys like "Seed: ", "Steps: ", "CFG scale: ", "Sampler: ", "Scheduler: ", "Denoising strength: " off it - the idea being that sampler settings come out of the workflow graph itself, no wires needed. That's genuinely the right instinct. But those A1111-style key names (with the trailing space) don't reliably line up with what current ComfyUI actually stores in its prompt data, so on modern builds those lines can come back as None. Treat the wired inputs as the source of truth and the sampler block as best-effort.

Inputs worth setting

• manual_input - a multiline string you type yourself; the top toggle (default on) places it above the generated block instead of below. Handy for adding a line like "upscaled 2x, denoised 0.4" that the node can't know.
• extra_info_input - a STRING that gets appended at the end, for whatever another node wants to say.
• lora_list - a flat list, three entries per LoRA: name, model strength, clip strength. This is exactly the triple of string outputs from the pack's Lora Loader with info, so the pack is clearly designed to chain lora_loader → image_info.
• vae_name - wire this from the pack's VAE Loader With Name so the caption says which VAE decoded the image.

The single output, Image information (STRING), goes wherever you want the recipe to live - a text/comment node to eyeball it, or folded into your save/display chain. Note that this node formats text; it doesn't write the workflow JSON into the PNG - ComfyUI already does that automatically, and the whole "workflow included" culture runs on it.

The honest assessment

For its narrow job - turning the wired inputs into a formatted block - it works, and the lora_list triple convention is neat. The INFO-based sampler section is the shaky part on current ComfyUI, and the misspellings ("posetive", "Strenght") are baked into the API surface, so copy the input names exactly when you search for them. If the sampler block matters to you more than the prompts, you might be better off with a maintained text/metadata node; if you just want checkpoint + prompts + LoRAs in a caption, this one is fine and has zero dependencies.

Install

Shared with the whole pack (four nodes, one repo):

cd ComfyUI/custom_nodes
git clone https://github.com/drago87/ComfyUI_Dragos_Nodes

Restart ComfyUI, or install via ComfyUI Manager by searching "ComfyUI_Dragos_Nodes". No extra Python packages, no model downloads. It's a small, unmaintained personal pack from late 2023 - rough around the edges, but the wiring for a metadata workflow is all here in one place.