Prompt Styler Viking Art

You type "a warrior standing on a cliff," pick Viking_Knotwork or Viking_Rune_Stone from a dropdown, and the node quietly wraps your subject in a paragraph of Norse art-direction - interlaced knotwork, weathered carved stone, animal-motif ornamentation - plus a matching negative. That's the whole pitch. It's a shortcut for the descriptive boilerplate you'd otherwise retype every single generation.

This is one node in wolfden's fork of the classic SDXL Prompt Styler, and its style list is 29 Viking-flavored looks: Viking_Knotwork, Viking_Runic_Inscription, Viking_Animal_Motifs, Viking_Ship_Design, Viking_Metalwork, Viking_Shield_Design, Viking_Mythology_Illustration, and more in that vein. If you're building Norse book covers, tattoo flash, game assets, or metal-album art, it saves you from memorizing what actually makes an image read as Viking.

How it actually works

There's no magic and no model here. Each style is a row in a JSON file with a prompt field containing a {prompt} placeholder and a negative_prompt field. The node drops your positive text into that placeholder and hands the assembled string back out. If you leave the negative blank, you inherit the template's negative; if you type your own, it gets combined with the template's. Pure string templating - it doesn't touch your GPU, call any API, or need a key.

Because it's just text assembly, everything depends on the checkpoint downstream. The styles were written for SDXL, and they lean on the tag-and-descriptor prompting that the SDXL lineage - Juggernaut, Illustrious, Pony and friends - was built to read. That's the context to keep in mind: this is an SDXL-era convenience, and it shines on SDXL-era models.

The inputs and outputs that matter

• text_positive - your subject. "a longship at dawn," "a bearded chieftain." Keep it short; the style supplies the flourish.
• text_negative - optional. Leave it empty to use the style's built-in negative, or add specifics you keep seeing (a stray modern object, a watermark).
• style - the dropdown. This is the node. Pick your Viking look here.

There's also log_prompt (No/Yes) - flip it to Yes and the fully assembled prompt prints to your ComfyUI console, which is the fastest way to see exactly what the template did to your text.

Two outputs: positive_prompt_text_g and negative_prompt_text_g. Wire the first into a CLIP Text Encode feeding your positive conditioning, the second into the encode feeding your negative. On an SDXL checkpoint you can feed that same G text into both the G and L boxes and be done.

Installing it

ComfyUI Manager: open Install Custom Nodes, search SDXL Prompt Styler (the wolfden one), install, restart.

Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/wolfden/ComfyUi_PromptStylers

Then restart ComfyUI. No pip install, no model downloads, nothing heavy - the whole pack is JSON and a bit of Python. After the restart you'll find the styler nodes under a Style Prompts submenu; click the Viking one and it drops into your graph.

Common issues

The first option does nothing on purpose. none byVikingArt is a passthrough - pick it and your prompt comes out unchanged. If the node "isn't styling," check you didn't leave it on none.

It's the wrong tool on Flux or Z-Image. On modern LLM-encoded models the negative half is inert at CFG 1, and the appended quality/style fluff gets read as a literal instruction rather than weighted tags - so the effect is muddy at best. Keep these stylers on the SDXL family they were written for.

Prompt soup. If you already stuffed your positive with lighting and quality tags, adding a heavy style on top dilutes everything. Let the style carry the atmosphere and keep your own text to the subject.

New node not showing? You skipped the restart. ComfyUI only scans custom_nodes at startup.