Seasonal Fashion Prompt Generator

The best thing about the Seasonal Fashion Prompt Generator is what it isn't: another LLM. No model downloads, no Python dependencies, no inference time, no GPU. It's a prompt generator that runs entirely on a pile of bundled CSV files, and that makes it the fastest way out of the blank-prompt problem I know of if you're making anime-style fashion shots on the Illustrious / Pony / NoobAI family.

You pick a season, flip a few toggles, hit "queue", and a complete comma-separated prompt comes out the STRING socket - something like winter, black turtleneck, pleated skirt, knee-high socks, walking, snow-covered street, golden hour. You wire that straight into your CLIP Text Encode (or a Positive Prompt node) and generate. For an SDXL-lineage anime model this is the right paradigm anyway: those checkpoints were trained on tag lists, so a structured tag prompt beats a sentence (prompt-engineering.md). The node is essentially doing what you'd do by hand, just faster and weirder.

How it works

The node ships with nine CSVs in its data/ folder - one each for seasonal fashion items, backgrounds, weather, times, poses, gaze directions, composition, body direction, and extra situational details. Every time it runs it randomly combines one item from whichever categories you left enabled, seeded by the seed input. That's the whole mechanism, and it's worth knowing one detail: the seed is clamped into a 16-digit range and passed to random.seed(), so the same seed always gives the same prompt. You can lock an outfit you like and only vary the seed.

The inputs that actually matter

There are 18 inputs but you'll touch maybe four. Each category toggle - top, bottom, one_piece, accessory, hat, shoes, socks, composition, gaze, pose, body_direction, weather, time - works the same way: set it to random (grab anything), off (skip it), or a specific value pulled from the CSV.

• season - spring / summer / autumn / winter, or random.
• background_toggle - on gives you a season-appropriate background plus weather/time/situation; simple swaps in a plain white/grey/black background, standing cut for clean character renders; off skips it.
• seed - your determinism handle. Change it for variation.
• wanna_shot - this is the joke. Set it to Yes and the node ignores most of the above and returns your outfit dropped into a cafe with coffee, holding a tumbler, straw inside the tumbler hardcoded in. It's a fun novelty and a good way to see what the fashion CSVs actually contain.

Install

It's a two-line install shared by the whole RS-Seasonal-Prompt-Generator pack (which also contains the Gemma TIPO node - both output a STRING, so you can chain them). Either use ComfyUI Manager and search RS-Seasonal-Prompt-Generator, or:

cd ComfyUI/custom_nodes
git clone https://github.com/raspie10032/RS-Seasonal-Prompt-Generator

Restart ComfyUI, done. This node specifically has zero dependencies and downloads nothing - the CSVs ship in the repo, so there's no first-run surprise and no model to babysit.

Gotchas worth knowing

Because there's no model, there's also no real failure mode - the worst that happens is a random season picks something you didn't want. Two practical notes: the output is plain English tag text, so don't expect it to auto-generate an image, and if you're on a natural-language base like Flux, a tag dump like this is the wrong prompt language entirely - this node is built for the tag-trained anime lineage. And if you hit the exact right seed, yes, you can replicate someone's exact prompt. That's a feature, not a bug.