Text Batch

Text Batch is the beefed-up cousin of Text Queue Processor. Same core job - feed ComfyUI a list of prompts and let it run through them one generation at a time - but with two upgrades that matter once your prompt list stops being three lines in a text box: it can read from a file, and it uses a block separator so a single "prompt" can span multiple lines.

That second part is the reason to pick this one. Plenty of good prompts are paragraphs, or have line breaks baked in. With a plain newline split, every line becomes its own generation and your carefully structured prompt gets shredded. Text Batch defaults its separator to ---, so you write each full prompt as a block and put a --- between blocks. Now a prompt can be as long and multi-line as you want.

How it works

Like the rest of aidec's queue nodes, it leans on a server-side counter instead of a real loop. Each time the workflow runs it emits the block at current_index, then advances. It's an output node, so it re-queues itself and marches through the whole list unattended. The difference from the plain queue processor is auto_stop - when it reaches the end it halts cleanly on its own rather than depending on something downstream to notice completed.

The inputs that matter

• input_mode - text to type your prompts into the node, or file to point it at a .txt on disk. File mode is the one you'll want once the list gets long or you're generating prompts elsewhere.
• text_file - the path, when you're in file mode.
• input_text - the multiline box, when you're in text mode.
• separator_type / separator - newline for one-line-per-prompt, or leave it as-is with separator set to --- for multi-line blocks. This is the headline feature; use it.
• start_index - where the walk begins (0 normally).
• auto_stop - leave it on so the batch ends itself.

The main output is prompt - note it's named prompt, not text like the queue processor - which you wire into CLIP Text Encode. Alongside it you get current_index, total, completed, and status. Route current_index into a filename or seed so your outputs stay traceable to the list.

Installing it

cd ComfyUI/custom_nodes && git clone https://github.com/aidec/Comfyui_TextBatch_aidec, then restart ComfyUI. (Or ComfyUI Manager → Install via Git URL with that repo link.) Nothing to download beyond the code itself - no models, no extra Python packages - and it shows up under the TextBatch category.

Where people get burned

• Separator mismatch. If your file uses blank lines between prompts but the node is set to split on ---, you'll get one giant "prompt." Match the separator to how your file is actually written - that's the single most common way this goes sideways.
• File paths. In file mode, give it a real, absolute path the ComfyUI process can read. A relative path or a typo'd folder just yields an empty batch. On a hosted/serverless ComfyUI, "on disk" means the server's disk, not your laptop's.
• start_index is sticky. After a full run, reset it to 0 before going again. It doesn't rewind on its own.
• One tab at a time. The counter is shared server state across the whole ComfyUI instance, so don't run this batch alongside another queue in a second window - they'll corrupt each other's position.

If Text Queue Processor felt too limited the moment your prompts got longer than a line, this is the one you actually wanted.