MD: Advanced Text Input

Plain text nodes are fine until you want a prompt that changes every run. AdvancedTextNode is the version that turns {red|blue|neon green} into an actual choice instead of a literal string - inline wildcards, seeded so you can reproduce a draw, with a pile of output wires that tell you exactly what happened.

If you've used wildcard systems before, the syntax will feel familiar: with wildcard_mode on, {option1|option2} picks one option at random, and you can switch the syntax to __option1|option2__ double-underscore style if that's what your collection of wildcard files speaks. Every draw is deterministic from a seed, so the same seed gives the same pick every time.

The inputs that matter

There's a lot of surface here, but a beginner only touches a handful:

• text - your prompt, template, YAML, whatever. Multi-line.
• wildcard_mode - the master switch. Off, and the text passes through byte-for-byte; on, and {a|b} patterns get resolved.
• seed and seed_mode - fixed for reproducibility, random for a fresh pick every run, increment for batch work where each run should advance the draw. The tooltip's suggestion of increment for batching is the right call.
• seed_list - a separate wildcard string ({1234|5678|9012}) that injects seed numbers rather than words, output on the selected_seed wire. Leave empty unless you're scripting seed sequences.

The rest is convenience: lowercase/uppercase (lowercase wins if both are on), strip_whitespace for cleaning pasted YAML, remove_extra_spaces (on by default, and you'll want it), and wildcard_syntax to pick the brace style.

What comes out

Six outputs, which is more than you'll usually wire up but useful when you need them:

• processed_text - the resolved string. This is the one that feeds your CLIP Text Encode.
• original_text - what you typed, untouched.
• seed_used and selected_seed - the actual seed that ran, and the value drawn from seed_list.
• text_length and wildcard_count - diagnostics, handy for debugging a prompt that's silently shorter than you think.

A nice workflow pattern: run with seed_mode set to increment, batch a few images, and the prompt picks (and seeds) step forward in lockstep so you can see what each option produced. Wire seed_used into your KSampler seed for full reproducibility.

Installing it

It's part of MD Nodes (MDMAchine/ComfyUI_MD_Nodes). ComfyUI Manager: search MD_Nodes, install, restart. Or the manual route:

cd path/to/ComfyUI/custom_nodes
git clone https://github.com/MDMAchine/ComfyUI_MD_Nodes.git
cd ComfyUI_MD_Nodes
pip install -r requirements.txt

Restart ComfyUI. The pack's requirements file pulls in a lot more than text processing needs (audio DSP, LLM integrations), but the one-liner is the supported path. Python 3.10+ and a current ComfyUI.

Common issues

People mostly trip on the obvious thing: typing {red|blue} and getting the literal string back. That's wildcard_mode being off - it's off by default, deliberately, so text passes through exactly as written. Flip it on and the pattern resolves. The other gotcha is expecting the double-underscore syntax to work while wildcard_syntax is set to curly_braces; pick one style and stick to it.