APNext Stuff

This is the "throw some things in the scene" node. APNext Stuff is the biggest of the dagthomas category builders - it's got dropdowns for everyday objects, gadgets, sci-fi tech, city fixtures, seasonal items, spooky props, and more, and it dumps your picks straight into your prompt. Need a random gadget on the desk, a piece of horror set-dressing, or 404 possible gizmos to roll from? That's this. The pack is the one people call the SDXL Auto Prompter, and Stuff is its junk drawer.

How it works

It's a menu system, not an AI. Nothing downloads, no API key, no network call. The pack ships huge curated word lists baked in - 404 gadgets, 227 everyday objects, 192 city items, 191 sci-fi, 192 scary things, plus office, seasonal, fantasy, and space sets - and each dropdown offers None, a specific item, Random (one pick), or Multiple Random (several). On run it gathers whatever the dropdowns resolved to, joins them with your separator, and tacks them onto your prompt. Seeded randomness means a roll you like is reproducible.

The inputs and outputs that matter

You won't touch most of the thirteen dropdowns on any given run - the point is to enable one or two.

• prompt (required, multiline) - your base text; picks get appended.
• separator (default ,) - the joiner.
• The category dropdowns - everyday, gadgets, scifi, city, space, fantasy, scary, office, icons, and the seasonal spring/summer/fall/winter - each None, a pick, Random, or Multiple Random.
• seed - fixes the rolls for reproducibility.
• attributes (boolean, default off) - adds descriptive adjectives when the pack's data attaches them to an item.
• string (optional) - extra text to fold in.

Outputs are two STRINGs: prompt (the assembled text) and random (the resolved-random version, so you can capture the exact roll). Wire either into a CLIP Text Encode.

How to install it

Via ComfyUI Manager: search comfyui_dagthomas or "SDXL Auto Prompter", install, restart. Manual install:

cd ComfyUI/custom_nodes
git clone https://github.com/dagthomas/comfyui_dagthomas
cd comfyui_dagthomas
pip install -r requirements.txt

Restart ComfyUI. The requirements file is beefy (transformers, decord, various provider SDKs) because the pack also carries a big set of LLM and vision nodes - Stuff itself needs none of them and runs purely locally.

Common issues & troubleshooting

Stuff is where over-enthusiasm bites hardest, precisely because the lists are so big. Set three dropdowns to Multiple Random and you can bolt a dozen unrelated objects onto one prompt - the model then tries to fit all of them and you get a cluttered, incoherent frame. Restraint reads better: one Random gadget or prop usually adds interest; ten fight each other.

Then the pack-wide caveat. The SDXL Auto Prompter name is literal - this emits comma-separated tags, ideal for CLIP-encoded models (SDXL, Illustrious, Pony). On 2026 LLM-encoded models (Flux, Z-Image, Anima) that comma soup works against the encoder, which reads the prompt as an instruction and drifts once you push past ~75-100 tokens. A wall of random objects is a subject-burial there. Keep it to SDXL-family models, or use a single pick as seasoning.

Last thing: keep the pack updated. It's large and actively developed, so an out-of-date copy is the most common cause of a load error - a git pull or Manager update has historically cleared the pack's install quirks.