Prompt Sequence Join

Prompt Sequence Join is the node that makes the rest of the pack worth installing. It takes two or more prompt sequences and merges them into one combined list - so you can run "four expressions × two outfits" as a single queued batch instead of hand-writing all eight prompts. That's the headline use case, and the README's minimal graph is basically this node: two Text nodes in, one text list out, straight into CLIP Text Encode.

But manage your expectations on the math. This is a zip, not a cartesian product. It pairs the first prompt of A with the first of B, the second with the second, and so on - it does not generate every combination. A 4-item list and a 2-item list give you 4 joined results (with the shorter list padded), not 8. If you genuinely want every cross product, this isn't that node.

How it works

Start with two text inputs; bump input_count and click update inputs to reveal more - up to 32 sockets, all fed by text_1 through text_32. Then decide how uneven lists behave:

• separator - the glue between parts, default ", ". It understands escapes: \n or /n become a newline, so ,\n gives you a comma-newline join for multi-line prompts.
• length_policy - longest (default) emits until the longest input is exhausted; shortest stops at the shortest connected input.
• missing_policy - what happens when a shorter input runs dry: blank (default) leaves the spot empty, repeat_last reuses that input's final value.
• skip_empty - when on (default), blank parts are dropped along with their separator, so you don't get ", , " gaps.
• max_items - caps the final joined count.

The README's example shows the default policy clearly. A is a1, a2, a3, a4, B is b1, b2. Output: a1, b1 / a2, b2 / a3 / a4. Switch missing_policy to repeat_last and it becomes a3, b2 / a4, b2 - B's last value carries the tail. That's the setting you want when a constant background or style should repeat across a longer subject list.

Output

text - one STRING list, the joined batch, wired into CLIP Text Encode like any other sequence output.

Install

ComfyUI Manager, search comfyui-prompt-sequence, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Localsmile/comfyui-prompt-sequence

Restart after either. No extra dependencies or model files.

The failure modes

The classic stumble is changing input_count and expecting the sockets to appear - they don't until you hit update inputs. Forgetting that is the top "why is it still showing two inputs" report. The second one is expecting cartesian behavior; when your batch comes out shorter than the number of combinations you imagined, re-read the zip explanation above. And the separator escapes only translate \n//n/\t - type the literal characters into the field, not an actual newline, or you'll get it right in the UI and wrong in the export.