🔧Text Encode for Sampler Params

Normal CLIPTextEncode gives you one conditioning per node. If you're comparing three or four prompt variants - testing wording, running an XY-style sweep, or generating a small set of related shots - the usual move is dragging out a separate text-encode node for each one and wiring them all up individually. Text Encode for Sampler Params collapses that into a single text box: write all your prompt variants in one place, separated by a line of dashes, and get them encoded together.

How it works

The default placeholder text in the node spells out the mechanism directly: "Separate prompts with at least three dashes\n---\nLike so." Put one prompt above a line of --- and another below it, and the node treats them as separate entries rather than one blob of text. Add more --- lines for more prompts. Under the hood it runs each chunk through the given clip and returns them as a set of conditionings rather than a single one - the multi-prompt equivalent of what you'd otherwise build with a stack of individual CLIPTextEncode nodes wired into a batch.

It lives in the pack's essentials/sampling category, which is the tell that it's built for exactly this kind of batch/comparison workflow rather than a single-shot generation - it's meant to feed whatever in your graph consumes a set of prompt variants for a per-prompt sampling run.

The inputs and outputs that matter

Just two inputs:

• text (STRING, multiline) - your prompt variants, one after another, each separated by a line of at least three dashes.
• clip (CLIP) - the text encoder, same as you'd wire into a normal CLIPTextEncode.

The output is CONDITIONING - but instead of one prompt's worth, it carries every dash-separated chunk you wrote, ready to drive whatever batch/per-prompt sampling setup you've built downstream.

How to install it

Via ComfyUI Manager: search "ComfyUI Essentials," install, restart ComfyUI. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/cubiq/ComfyUI_essentials

No extra model downloads - it's a text-processing wrapper around whatever CLIP encoder you already have loaded.

Common issues

Fewer than three dashes doesn't split. The placeholder text is specific about this for a reason - a stray -- or a single - on its own line won't be treated as a separator, and your "two prompts" will get encoded as one run-on block instead. Stick to a clean --- line.

Smart-quote or em-dash autocorrect. If you're pasting prompt text from a word processor or a site that auto-converts hyphens, double-check the separator line actually contains plain ASCII hyphens (-) and not an em dash or other lookalike character - those won't match the split.

Expecting a single conditioning out. This node's whole purpose is producing more than one conditioning from one text box. If your downstream node only expects a single CONDITIONING input, you may need to route through something that selects or batches appropriately rather than assuming a 1:1 wire like you'd get from plain CLIPTextEncode.