textOutput

textOutput is the text-world equivalent of a Save Image node: it takes a string, shows it to you, and - if you ask - writes it to disk as a real file. It's the node you reach for when you're generating prompt variations or positive/negative pairs and you want them preserved as .txt, .md, or .csv files instead of living only in a workflow JSON.

The workflow where this earns its keep: you've got a textConcat or a pipeLoader's prompt string, and you want an audit trail of exactly what got sent to the sampler - including the resolved wildcards, which the static text boxes don't show you. Point this at the end of your text chain, flip it to Save, and every run leaves a timestamped file you can diff later.

How it works

It's an output node with two modes via text_output:

• Preview - renders the text in the node's UI area. No file.
• Save - writes the text through the pack's shared save machinery, with all the filename goodies.

When saving, you get output_path (defaults to ComfyUI's output folder), save_prefix (supports the %date:yyyy-MM-dd-hh-mm-ss% syntax and %widget%-style value injection), number_padding for the counter, and file_type - txt, md, rtf, log, ini, or csv. The overwrite_existing toggle controls whether a repeated run rolls to a new number or replaces the file.

The text output passes the string through untouched, so you can tee it into another node (say, a text concat that builds a filename from it) while still saving.

Inputs and outputs

• text_output - Preview or Save. The one you'll actually flip.
• text - the STRING to display/save. Usually wired from a concat or prompt node.
• text - the same STRING as a passthrough output.

Installing it

It's part of the tinyterraNodes pack:

cd ComfyUI/custom_nodes
git clone https://github.com/TinyTerra/ComfyUI_tinyterraNodes.git

or ComfyUI Manager → tinyterraNodes. No dependencies, no models.

Gotchas

The classic complaint is the other direction: people expect this to load text from a file and it only saves. If you want to read a file into a string, that's a different node (the pack doesn't ship a file-reader; WAS Node Suite or a core load node handles that). Also, csv output is a single cell, not a nicely formatted table - it's for when downstream tooling wants a plain value, not for spreadsheets. And remember the %date% prefix parsing is done at save time, so two runs in the same second with overwrite_existing off will collide into a (1)-style suffix rather than a separate timestamp. Minor, but it's caught me before.