AI2Go Prompt Batch

ComfyUI doesn't have a real for-loop. The classic workaround for "generate the same image with twenty different prompts" is the Load Image Batch + increment index trick: queue N runs by hand and let a counter walk a list. This node is the text analog. Paste in a JSON list of prompts, queue N runs, and each run emits one prompt from the list. It won't change your life if you generate one image at a time - it absolutely will if you've ever sat there queueing the same graph ten times to test variations.

How it works

The prompt list lives in a hidden prompts_json field that's the source of truth, kept in sync with a row editor on the node itself. Each row is a positive and negative text box side by side; ➕ Add Prompt appends a row, 🗑 removes one, and the ⠿ handle drag-reorders. The node auto-grows and auto-shrinks to fit. An index counter picks which prompt fires this run, and the front-end advances it by 1 after each queued run - so a queued batch walks 0, 1, 2… down the list. It's deliberately immune to the "Widget Value Control Mode" setting, which is a classic ComfyUI footgun that quietly freezes widget state mid-batch.

The practical loop looks like this:

1. Build your list in the rows editor.
2. Click 🔍 Check for prompts - it validates, counts N, and resets the index to 0.
3. Set ComfyUI's queue count to N and run.

Two toggles worth knowing: reset_index_at_batch_start (on by default) zeroes the index when a new batch is queued so every batch starts from the first prompt, and delete_empty_prompts (on by default) drops blank rows so they never break a run or skew your count.

The JSON

Each entry is an object with a required positive and optional negative:

[
  { "positive": "a red fox in the snow, highly detailed, 8k", "negative": "blurry, watermark" },
  { "positive": "a neon-lit city street at night, cinematic" }
]

The importer is forgiving: a bare string counts as positive-only (["a red fox", "a neon city"]), prompt works as an alias for positive, and the array can be wrapped as { "prompts": [ ... ] }. There's also a json_in STRING socket - wire a text/primitive node holding a JSON list here, click 📥 Read from JSON, and it appends into your rows. Note the socket is ignored at run time: the rows you edit are what actually run.

The outputs

positive and negative - feed them into your CLIP text encode. Plus index, the 0-based index used this run. Wire that into a SaveImage filename and each output file records which prompt made it. That detail is also why this node pairs well with the pack's AI2Go Save Metadata (Civitai) nodes - their trace reads the batch node's current index and resolves the exact line that ran, so the PNG metadata matches the image.

Gotchas

• Overshoot the list (queue more runs than prompts) and the index clamps to the last prompt instead of erroring - arguably right, but worth knowing so you don't silently regenerate your final prompt N times.
• With delete_empty_prompts off, an empty positive raises a "positive is empty" error at run time. Leave it on.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/Little-God1983/ComfyUI-AI2Go-Utils

Restart ComfyUI. Pillow is the only dependency, no model downloads. And the usual pack disclaimer: published for reference, issues disabled, GPL-3.0 - you can fork it, you just can't expect the author to hold your hand.