Text Index Switch 4

Here's the thing most tutorial workflows hide from you: random prompt selection isn't one node, it's two. You need something to generate the index and something to switch on it. Text Index Switch 4 is the AUN pack's answer to that - it folds the index generator and the text switcher into a single node, with four selection modes, and it's the node I'd use for anything prompt-randomization related. It saves you a node and a wire, and those are the kinds of savings that add up across a hundred workflows.

What it does

It combines index generation (Select / Increment / Random / Range) with up to 20 text slots. You set a mode, and the node produces the index internally - no separate Random/Select INT node required - then outputs the matching text plus the label and the active index. The README's "how to randomly select a prompt" workflow uses the two-node setup with AUNRandomIndexSwitch; this node is the one-node version of that exact idea.

Inputs that matter

• mode - the selector: Select (fixed index), Increment (cycle through the range, great for batch runs), Random (random within range), Range (pick from a list like 1,2,5-8,12).
• minimum / maximum - the inclusive bounds for Random/Increment modes, clamped to your slot count.
• index - the current/target index (what Select mode uses, and what Increment cycles).
• slot_count - how many text slots are active, 1–20.
• range - the comma-separated index list for Range mode, e.g. 1,2,5-8,12.
• text1–text20 - the prompt slots, multiline, with dynamic-prompt support.

Outputs: text, label, and index (the active index - wire it into other controllers or filename builders).

How it works

The state lives in the node itself - it remembers its current index for Increment mode and cycles through the range, so consecutive runs step through your prompts in order. Random mode uses a proper system-random source, and Range mode parses the list (including 5-8 style spans) into a valid set, cycling through it predictably. There's also an IS_CHANGED-style re-execution trigger, so Random/Increment/Range modes actually re-run each time instead of getting cached by ComfyUI - the silent killer of "why is my random prompt stuck on the same one?" - is exactly what this avoids.

The label works like the other switches: connected nodes contribute their titles, otherwise the slot's text stands in. And because slot_count genuinely bounds the active slots, a maximum beyond your count just clamps instead of crashing.

Which of the four switches do you want?

• V1 TextIndexSwitch - bare switch, cleanest when you're feeding the index from elsewhere.
• TextIndexSwitch3 - adds the first-line-as-label convention and an index output; great for hand-organized prompt cards.
• TextIndexSwitch4 - index generation built in, the all-in-one. This is the one for "give me random prompts" with the fewest moving parts.

For most people, the honest answer is: start here. One node, four modes, no extra wiring.

Installing

Ships in the AUN ComfyUI Nodes pack:

cd ComfyUI/custom_nodes
git clone https://github.com/loz2754/AUN-ComfyUI-Nodes

Restart ComfyUI, or use ComfyUI Manager (search "AUN"). No models, no heavy deps beyond the pack standard (auto-installed by Manager, or pip install -r custom_nodes/AUN-ComfyUI-Nodes/requirements.txt for manual clones).

One tip before you run it: if you're using Increment mode for a batch, set maximum to exactly your slot count and keep minimum at 1 - otherwise the cycle range and the slot list can disagree, and you'll wonder why slot 12 never gets picked.