Remove Empty Lines And Leading Spaces 📝

Every prompt you copy out of a shared workflow or a website comes with baggage: blank lines, stray indentation, a leading space that quietly changes how the CLIP text encoder tokenizes the first word. This node is the lint pass. It takes a block of text, strips the cruft, and hands back something clean enough to paste straight into a prompt or a LoRA training caption file.

It's part of the Painting Coder Utils pack from jammyfu, a small collection of image-and-text helpers that all share the PaintingCoder:: namespace and the same no-frills install. Nothing here calls an API, downloads a model, or needs a key - it's regex and string work end to end.

How it works

The node takes your text and runs two independent cleanup passes, each controlled by its own toggle:

• remove_empty_lines_option (on by default) drops blank lines, collapses runs of empty lines down to one, and trims leading/trailing newlines.
• remove_leading_spaces_option (on by default) strips leading spaces on every line and then trims each line's leading/trailing whitespace.

So this mess:

    Hello World
    
     This is a test
    
    
    Multiple empty lines above

becomes, with defaults:

Hello World
This is a test
Multiple empty lines above

Both toggles default to on, so out of the box it does the full cleanup. Turn one off if you care about that specific thing - say you're feeding a markdown-ish block where indentation means something.

The inputs and output that matter

• text - the multiline string you're cleaning. This is the only required input.
• output_type - String or List. String joins the cleaned lines back together with newlines; List hands you an actual list you can feed to nodes that expect a list of strings (like a batch text processor).
• Output: output - the cleaned text (or list).

One thing to know: this node doesn't display anything. If you want to see what it produced, run the output into a Show Text Plus or any display node. That trips people up more than you'd expect - you wire it in, queue the graph, and "nothing happens," because the cleaned text is just silently waiting on the output port.

Install

Use ComfyUI Manager and search for Painting Coder Utils, or clone it manually:

cd ComfyUI/custom_nodes
git clone https://github.com/jammyfu/ComfyUI_PaintingCoderUtils.git

Then restart ComfyUI. The pack's requirements.txt lists numpy, Pillow, torch, and gradio - all things a working ComfyUI install already has, so there are no model downloads and no heavy surprises. If you're loading an old workflow that uses the pre-0.3.0 class names, jammyfu ships a workflow fixer tool in docs/fix/workflow_fixer.html to repair the JSON - worth knowing before you assume the node is broken.

Where you'll actually use it

• Cleaning pasted prompts before they hit a CLIP Text Encode, especially when the source was a multiline block with junk whitespace.
• Formatting caption files for LoRA training, where a stray blank line or leading space can mess up the dataset parsing.
• Normalizing text that came out of another node - several of this pack's own text nodes produce multi-line output you'll want tidied before reusing.

It's a boring node, and that's the compliment. If you've ever stared at a workflow and wondered why "cat" with a leading space wasn't matching your tags, this is the fix.