Text String (Mango)

Text String (Mango) is about as simple as a node gets: it's seven labeled text boxes - face, eyes, hands, feet, shoes, gen1, gen2 - where each box comes out on its own STRING output, unchanged. Type in the face box, and the face output carries that text; leave the hands box empty and the hands output is empty. It's a passive text multiplexer, not a prompt builder.

Why would anyone want that? If your workflow does regional or character-detail conditioning - separate CLIP Text Encode nodes for face details, hands, clothing, and so on - a single node with named, color-coded sections keeps the graph tidy and gives you seven labeled outlets to wire into different text encoders instead of a pile of free-floating "String" nodes. The labels are suggestive, not enforced: nothing stops you from putting scenery text in face. They're just convenient names.

How it works

The mechanism is literally a pass-through. The execute function takes all seven inputs and returns them in the same order, as-is. There is no concatenation, no formatting, no default values injected. The only real logic is in the node definition: face is required (defaulting to an empty string), and the other six are optional. That's the entire implementation, and it's fine - the node's job is organizational, not computational.

Inputs and outputs

• face (required) - the one box you must populate; empty string is fine.
• eyes, hands, feet, shoes, gen1, gen2 (optional) - six more boxes, each defaulting to empty.
• Outputs - face, eyes, hands, feet, shoes, gen1, gen2, each a STRING mirroring its input.

Every input has a matching output of the same name; there's no combined output, so plan on wiring each one where it needs to go.

Installing it

Text String (Mango) is part of the Mango Node Pack:

cd ComfyUI/custom_nodes
git clone https://github.com/mang01010/MangoNodePack

Restart ComfyUI, or install "Mango Node Pack" from ComfyUI Manager. The README's only extra dependency is pip install safetensors.

Gotchas

The honest review: this is a very thin utility, and whether it earns its place in your graph depends on how you work. If you already use a "Show Text"-style workflow with loose string nodes, switching to named boxes is a small clarity win. If you were hoping it would build a prompt (concatenate face + eyes + hands into one string), it won't - that's what Prompt (Mango) and Prompt /w Embedding are for. One small trap: because all seven outputs exist even when their boxes are empty, wiring an empty box into a text encoder produces empty conditioning - which is fine (an empty prompt is a valid prompt), but you'll wonder why a region looks uncontrolled until you realize the box was never filled. Use it for what it is: seven labeled lanes for text, nothing more.