Load Text Line

Load Text Line is the file version of a prompt switcher. You keep a text file of prompts (or negative prompts, or styles - one per line), and this node hands you one line at a time: the whole file, a specific line by index, or a line drawn from a seed. Feed the index from a counter and you step through a list run after run; feed it a seed and every run pulls a different random line you can reproduce. It turns a plain .txt into a batch-run schedule without any picker UI or extra files.

It lives in WAS Node Suite (menu path WAS Suite/IO) - the big MIT toolkit whose v3 rewrite pushed past 450 nodes. If you've ever used its Text Random Line or the wildcard machinery, this is the more deliberate cousin: you control which line, not just "a random one".

How it works

The file menu scans ComfyUI's input and output folders (plus anything under paths.allow_read in config) for .txt, .csv, .tsv, .json, .jsonl, .md, .yaml, and .yml files, tagging entries [input] or [output] so two files of the same name stay distinguishable. Drop a new file in and it shows up within about five seconds. A file that's since been deleted gives empty text and a log note rather than killing the prompt - handy in long batch runs.

Three modes decide what the main line output carries:

• file - the whole file, lines joined back together. That's the one that feeds a prompt written across several lines.
• index - the single line at index, counting from 0. Negative indexes count from the end (-1 is the last line), so it's safe to reach for the last line of a file whose length you don't know.
• random - one line drawn by seed. Same seed + same file = same line, every time. Worth knowing the seed pins a position, not a line: add or remove a line in the file and the same seed silently gives you different text.

skip_comment_lines is on by default, which drops any line whose first non-space character is # before anything is counted or numbered - so you can keep notes at the top of a prompt file without polluting your runs.

What comes out

The node hands you more than the one line:

• lines - every line as an ARRAY, ready for Text List Get/Slice/Length.
• text - the whole file as one string whatever the mode, for something like Text Find and Replace that wants the file as a whole.
• line_count - how many lines survived comment filtering (0 for a file that couldn't be read).
• resolved_index - which line number line actually came from after wrapping or clamping. When a counter is driving the index, this is your sanity check.

The out_of_range setting matters more than it looks: wrap makes line 5 of a 3-line file read line 2 (good for cycling forever off a climbing counter), clamp sticks at the ends, empty returns nothing and keeps the graph running, and error stops the prompt - the right choice when running off the end means something upstream genuinely broke.

Installing it

Standard WAS install. ComfyUI Manager → search "WAS Node Suite v3", or clone it:

cd ComfyUI/custom_nodes
git clone https://github.com/WASasquatch/was-node-suite-comfyui.git

Then restart ComfyUI. Needs ComfyUI 0.14.0+ and Python 3.10+; the pack adds no pip dependencies.

Where people get burned

The classic gotcha is the comment filter silently renumbering your file. If you're driving index from a counter and the file has # headers, every index after the header is off by however many comments you wrote - resolved_index output is there precisely so you can see it. The second is forgetting the file lives in input or output folders: for a file anywhere else on disk, the suite's Load Text File takes a typed path instead.