◈ Artists

Pony Diffusion's author stripped artist tags from its training data deliberately. Illustrious kept them, and it's one of the biggest reasons artist-tag prompting became a real style lever in this ecosystem instead of a novelty - NoobAI in particular leans on artist tags specifically to get a consistent aesthetic, and the community built a dedicated 16,000-artist style explorer just to help people find ones that actually work. This node is the pack's front end to that vocabulary: five artist slots, deliberately graded in influence.

Five slots, deliberately unequal

Read the tooltips on the five fields and the pattern is explicit: Artist (Base) "heavily influences the prompt," down through Artist (Quaternary) and Artist (Quinary), both labeled "influences the prompt very little." This isn't five artists blended flat - it's a primary style with up to four progressively lighter accent influences, which lets you do something like "mostly wlop's linework and color, with a little kantoku mixed in" instead of a 50/50 mash that reads as neither.

Weighed Average changes how the five combine into one result - as with the Characters node, the exact blending math isn't published in the schema, so treat it as worth A/B-testing on your specific artist combination rather than something you can predict from the tooltip alone. Format Tag wraps output in escaped parentheses for weight emphasis, the same convention used across the rest of the pack.

Why a typed artist name sometimes does nothing

The most common reason an artist tag produces no visible effect isn't a typo - it's under-representation in the training data. The rough rule of thumb that applies to characters (roughly 200+ works on Danbooru before a model reliably "knows" something) applies to artists too. A correctly spelled, real artist tag for someone with a thin Danbooru presence will still read as close to generic, because the model simply didn't see enough of their work to learn a distinct style from it.

The inputs and outputs that matter

• Artist (Base/Secondary/Tertiary/Quaternary/Quinary) - five STRING fields with genuinely different weight by design, not five equal inputs.
• Weighed Average - off by default.
• Format Tag - on by default.
• prefix / suffix - splice text around the assembled artist tags.

Output is a single ARTIST PROMPT string - combine it with IllustriousCharacters and your scene description into your main prompt.

Installing it

ComfyUI Manager: search "EasyIllustrious," install, restart. Manually:

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

No downloads - this node's field values are free text against your checkpoint's own learned vocabulary, not a bundled asset you need to fetch.

Common issues

An artist tag you're confident is spelled correctly still does nothing visible. Check the artist's actual Danbooru representation before assuming the node or your spelling is wrong - thin representation in training data is the more common cause than a typo. Cross-checking against Danbooru's own tag wiki, or a community style-lookup tool, is faster than guessing.

Trying to blend five artists evenly and getting a muddy, indistinct result. That's expected given how the slots are weighted - Base dominates by design. If you actually want closer to equal influence between two or three styles, put your strongest pick in Base and treat the rest as accents, or drop to fewer named artists and lean harder on Weighed Average.

Unsure if an artist tag is even real Danbooru vocabulary or something invented. The node accepts any string - it doesn't validate against a known-artist list before encoding, so a nonsense or misremembered name will silently produce nothing rather than erroring.