Fat Mex Prompt Batch

If you're generating variations for a content pipeline, the boring part is typing the same thing into a prompt box again and again. This node encodes a whole list of prompts in one pass: one prompt per line, all of them CLIP-encoded up front, one shared negative, and the raw strings passed through so you know what you generated. It replaces CR Prompt List + CLIPTextEncode with a single node that does the encoding step too.

How it works

Type (or paste) multiple prompts, one per line. The node:

1. Splits on newlines and strips empties.
2. Applies the optional prefix and suffix to every line - useful for injecting a shared style block ("cinematic lighting, 85mm") or a shared camera directive without retyping it ten times.
3. CLIP-encodes each line as a positive conditioning.
4. Encodes the single negative_prompt once, shared across the batch.
5. Hands back everything: positive_batch (a list of conditionings, one per prompt), negative (the shared one), count (how many prompts), and prompt_list (the raw strings, for display or logging).

dynamicPrompts is on for the prompt field, so wildcard-style expansion works if you have that convention in your pipeline.

What wires where

This is the part that needs a beat of explanation, because the outputs are conditionings, not images. positive_batch is a list, so you need a sampler that can consume a conditioning list (or a loop/batch node), or you pull individual items out and feed each to a sampler. The count and prompt_list outputs are there so a downstream node can iterate over the batch and label what's what.

The pack's own example workflows actually show the simpler sibling pattern - CR Prompt List feeding the Fat Mex Content Sampler one prompt at a time. Fat Mex Prompt Batch is the version you reach for when you want the encoding done once, up front, instead of inside each sampler pass.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/FatMex/ComfyUI-FatMex-Nodes.git FatMex-Nodes
pip install -r FatMex-Nodes/requirements.txt

Or ComfyUI Manager → search "Fat Mex Nodes" → Install → restart. No model downloads needed - this node just needs a CLIP, which it takes as an input.

Where people get burned

• Nothing appears on screen. Remember the contract: this node produces conditionings, not images. You still have to feed them into a sampler and decode. If you ran it and "nothing happened," that's expected - it's a middle node, not an output.
• Empty prompt field throws. It raises a clear "No prompts provided" error, which is the one failure mode that actually errors rather than silently passing.
• Shared negative, shared prefix. The negative applies to every image, and prefix/suffix hit every line. That's the design - but if you wanted per-line negatives, this isn't the node (there's one negative_prompt box for the whole batch).