APNext Vehicle

Making car content and tired of typing model names? APNext Vehicle is a dropdown stuffed with them - 308 modern cars, 147 classics, 80 supercars, 103 vehicle types (boats, blimps, backhoes, the lot), plus a fun list of 38 famous movie cars like the DeLorean and Herbie. Pick one, or roll Random, and it appends the term to your prompt. It's part of the dagthomas family the community calls the SDXL Auto Prompter, and it's the most specialized of the category nodes - genuinely handy if vehicles are your thing, pointless if they're not.

How it works

It's a menu, not a model - nothing downloads, no API key, no network. The pack ships curated vehicle lists baked in, and each dropdown offers None, a specific vehicle, Random, or Multiple Random. On run it grabs whatever the dropdowns resolved to, joins them with your separator, and appends them to your base prompt. Seeded randomness means a roll you like is reproducible.

The inputs and outputs that matter

• prompt (required, multiline) - your base text; picks land after it.
• separator (default ,) - the joiner.
• The five dropdowns - cars (modern), classic_cars, fast_cars (supercars), types (categories of vehicle), famous_cars (movie/TV rides) - each None, a pick, Random, or Multiple Random.
• seed - fixes the random rolls.
• attributes (boolean, default off) - adds descriptive words when the pack's data attaches them to a term.
• string (optional) - extra text to fold in.

Two STRING outputs: prompt (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

ComfyUI Manager: search comfyui_dagthomas or "SDXL Auto Prompter", install, restart. Or manually:

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

Restart ComfyUI. The requirements list is heavy (transformers, decord, provider SDKs) because the pack also ships a big set of LLM and vision nodes - Vehicle uses none of them and runs purely locally.

Common issues & troubleshooting

The specific gotcha for Vehicle: a named car is only as good as the model's training on it. Common models - a Mustang, a 911, a DeLorean - usually render right. Obscure trims and the deep-catalog classics are a gamble; a checkpoint that never saw a "1905 Stanley Steamer" will just invent an old-timey car and hope. The types dropdown (generic "sports car", "pickup truck") is the more dependable steering when you don't need a specific badge. And branded trademarks can render inconsistently or with warped logos - that's the model, not the node.

Then the pack-wide caveat, because it decides whether this node helps you at all. The SDXL Auto Prompter name is literal: it emits comma-separated tags, which suit CLIP-encoded models (SDXL, Illustrious, Pony). On 2026 LLM-encoded models (Flux, Z-Image, Anima) the encoder reads your prompt as an instruction, so "a red 1967 Mustang parked on a wet street" beats Mustang, 1967, red - and firing Multiple Random across every dropdown buries your subject past the ~75-100-token mark where those models drift. Use one pick, and prefer weaving it into a sentence on LLM models.

Finally, keep the pack updated. It's large and actively developed, and a stale copy is the usual cause of a load error - a git pull or Manager update has historically been the fix.