Skeba Batch Text (Prompt Loop)

Skeba Batch Text (Prompt Loop) is the text twin of the folder loader: you paste a bunch of prompts into one node, it splits them into a real ComfyUI list, and any batch-aware node downstream runs once per prompt. It's the fastest way to churn through a set of prompt variations - a scene description written five ways, ten character concepts, a shot list for a video chain - without building five nearly-identical branches.

It lives in the ComfyUI-H3-Reference-Library pack (a.k.a. ComfyUI-Minimax-H3-Reference-Library), under Skeba AI Nodes - Utilities, and it's the descendant of the original SkebaPromptLoopNode from the old batching pack. The class name is preserved on purpose, so workflows saved against the older node load into this one unchanged.

How it works

The mechanism is a split() plus a strip. It takes your text (a multiline STRING), splits on delimiter (default \n), trims whitespace off each entry, and if skip_empty is on (it is by default) drops blank lines. That's it - no parsing of prompt syntax, no magic.

The four outputs are where you pick your adventure:

• prompts_list - a standard STRING list. Wire this into nodes that execute per item. This is the batch path.
• count - how many prompts came out, useful for a loop counter or a progress display.
• prompt_numbers - an INT list, one-based (1, 2, 3...). Handy if you want each output tagged with its number.
• prompt_batch - the same prompts repackaged as a custom SKEBA_PROMPT_LIST value. This one exists for Skeba Select Prompt From Batch, which consumes that exact type.

Most people want prompts_list. The prompt_batch output only matters when you're doing indexed selection instead of full iteration.

Why you'd reach for it

Video workflows especially. H3 and other video models take a while per clip, and the difference between "queue these five prompts" and "reload the workflow, edit the text box, queue again, repeat" is the difference between going to bed and babysitting the machine. It also composes with the image folder loader in the same pack: list of images × list of prompts, one queue, a stack of outputs.

Installing it

ComfyUI Manager - search "Skeba" or "H3-Reference-Library" - or:

cd ComfyUI/custom_nodes
git clone https://github.com/nikaskeba/ComfyUI-H3-Reference-Library

Restart ComfyUI. No extra Python dependencies, no model downloads.

Where people get burned

• Every prompt runs the whole downstream graph. That's the deal with list batching, but it bites people on video: five prompts is five full generations, not five cheap variations. Budget the queue accordingly.
• The delimiter is a literal string. Default is a newline, so one prompt per line. Change it to | and it splits on pipes. If you change it, remember skip_empty still applies.
• Turn skip_empty off and you get blank strings in the list, which some nodes choke on or turn into empty generations. Leaving it on is the right call for nearly everyone.