CreaPrompt Multi Prompts

Every CreaPrompt node hands you a multi-line string when you bump Prompt_count above 1, but a multi-line string isn't a batch - a batch is a list of prompts with a matching list of seeds, and that's exactly what CreaPrompt Multi Prompts exists to build. It's not a generator itself; it's the glue that turns the randomizer's output into something you can run N images against, each with its own reproducible seed.

This is the node the pack's shipped workflows all use. Open one_node_workflow.json or four_nodes_workflow.json in the WF folder and you'll see the pattern: CreaPrompt node → CreaPrompt Multi Prompts → a sampler that accepts a prompt list. The pack's own examples treat this node as part of the standard rig, not an optional extra.

The inputs

Four inputs, all simple:

• Multi_prompts (multiline STRING) - the lines you want as separate prompts. Feed it a CreaPrompt node's prompt output directly, or paste your own list. Each line becomes one prompt in the output list.
• seed_base (INT, forced input) - the starting seed. The node generates one seed per prompt as seed_base + 0, seed_base + 1, ... so every line in your batch gets its own reproducible seed, in order.
• prefix and suffix (multiline STRING, both optional) - a fixed string prepended and/or appended to every line, comma-joined. Want "masterpiece, best quality" on every prompt? Drop it in prefix once instead of editing N lines. The suffix slot is handy for a consistent quality tag or a shared negative-ish constraint.

The outputs

• prompt (STRING, list) - one entry per input line, in order. Wire this into a sampler or node that consumes a prompt list for batched generation.
• seed (INT, list) - the matching seed list, one per prompt, so prompt i and seed i travel together. Same length, same order - that's the synchronization you'd otherwise have to hand-roll.
• prompt_debug (STRING, not a list) - a single human-readable dump of every line it processed (the lines are prefixed with an arrow for readability). Preview it with a text/string display node when something looks off and you want to see exactly what got built.

How it works

The source is almost suspiciously short: strip the input, split on newlines, apply prefix/suffix, build the seed list, done. No randomness, no CSV involvement - it's pure text plumbing, and that's why it's reliable. The only real logic is the seed arithmetic, and it's worth knowing it's seed_base + index, so the order of your lines matters if you care which seed maps to which prompt across runs. Keep lines stable and the batch stays reproducible.

Installing it

Same as the pack - Manager (search "CreaPrompt") or:

cd ComfyUI/custom_nodes
git clone https://github.com/tritant/ComfyUI_CreaPrompt

Restart. No dependencies, no downloads, offline. This node has nothing heavy about it.

Common issues

• A prompt list where you expected a string (or vice versa). The prompt and seed outputs are true lists; some downstream nodes want a single string. That's not a bug, it's the point - check you're wiring the list outputs into a node that accepts lists.
• Seeds don't match your expectations. It's literally seed_base + line index. If you reorder lines, the seed-to-prompt mapping shifts with them.
• Trailing blank lines. A stray empty line at the end of Multi_prompts becomes an empty prompt in the list. The node strips whitespace but doesn't drop empty lines, so clean input stays clean output.