Power Artist Loader
Load your whole artist library into one prompt — Power Artist Loader
- text
If you've ever stared at the prompt box trying to remember exactly how you spelled that one artist's style keywords, this node is for you. Power Artist Loader is the flagship node of the small ComfyUI-Power-Artist-Loader pack, and it does one thing: turn a pick-and-click list of artists into a ready-made prompt string. It's the rgthree Power Lora Loader idea - stacked slots, per-slot toggles and weights - applied to artist names instead of LoRAs. The README says so outright, and the lineage shows.
Why you'd reach for it
You run anime-style tag models - Illustrious, NoobAI, Pony, anything in the SDXL/Danbooru lineage - and you've accumulated a vocabulary of style triggers. "Akira Toriyama style, anime, dragon ball." "WLOP:0.5, Pablo Picasso:0.8." Power Artist Loader keeps those in a plain CSV file, shows them in a dropdown, and resolves whatever you pick to its keyword expansion. It's a library, not a magic prompt generator. If your workflow is more "Flux 2 with a written description" than "tag soup", this is probably not the tool you need - the whole thing is built around comma-separated tag syntax.
How it works
Here's the odd bit: the node has no input sockets. Its schema's required and optional inputs are both empty - the only plumbing is a single STRING output named text. All the state lives in the node's widgets plus a artists.csv file in the pack folder.
On every execution it re-reads that CSV (semicolon-separated: name;keywords;imagefile), walks its internal artist_N widget slots (each carrying an on flag, the selected artist, and a strength), looks up each enabled artist's keywords, and formats them. Strength 1.0 passes through bare; anything else gets wrapped in the SD weight syntax - (keywords:1.20). Then it joins everything with commas, strips the spaces after the commas, and (if you typed anything into the base text field) appends it. One clean STRING comes out the other end.
The widgets are where the fun is, straight from the README:
- Add Artist - adds a slot.
- Dropdown - pick an artist; hover for a live style preview if you've dropped a thumbnail into
images/. - Weight - ±0.05 arrows, click-to-type, or drag. Range −1.00 to 3.00, negative weights included.
- Toggle All and the per-slot circular toggle - mute everything or one at a time.
- Right-click - quick set 0.5 / 1.0 / 1.5.
Where the output goes
The text output is a plain string, so it feeds any prompt path: a string concatenator, then into a CLIP Text Encode positive prompt. That's exactly how the pack's own example workflow wires it. One important caveat: the (keywords:1.2) weight syntax is CLIP-era. On LLM-encoder models like Flux 2, Anima or Z-Image those weights are silently discarded - your artists still appear, but the strength numbers do nothing. Save them for your tag-lineage checkpoints.
Installing
ComfyUI Manager is the easy route - search "ComfyUI-Power-Artist-Loader" and install. Or the manual way:
cd ComfyUI/custom_nodes
git clone https://github.com/hanL34/ComfyUI-Power-Artist-Loader.git
Then restart ComfyUI. Good news: there's no requirements.txt and nothing heavy to download. The only imports are numpy, Pillow, aiohttp and aiofiles, which ComfyUI already ships, and the whole thing is local - no API keys, no model files, no network calls.
Making it yours
Your artist library is just custom_nodes/ComfyUI-Power-Artist-Loader/artists.csv:
Akira Toriyama;akira toriyama style, anime, dragon ball;toriyama.jpg
厚涂风格;(WLOP:0.5),(Pablo Picasso:0.8),(Greg Rutkowski:1.1);houtu.jpg
Semicolons between fields, commas inside the keywords, UTF-8. Drop a matching 200–400px JPG into images/ and hover-preview kicks in. Edit the CSV (or use the right-click Artist Manager window) and the loader picks it up on the next execution.
Common issues
- Nothing loads after editing artists.csv - the frontend caches; refresh the page, the node re-reads the file on every run.
- Preview image missing - the filename in the CSV has to match a file in
images/exactly. - Weights seem ignored - you're almost certainly on an LLM-encoder checkpoint. See above.
- Rows parse wrong - you used commas as the field separator. It's semicolons.
This is a small, MIT-licensed hobby pack from a solo author (hanL), maintained in fits - changelog bursts in late 2025, a PR merged in May 2026 - and it has essentially no community footprint to lean on. It works, it's lightweight, and for building a repeatable anime-style artist toolkit it's genuinely handy. Just don't expect hand-holding.
Inputs (0)
No inputs
Outputs (1)
| Name | Type | Description |
|---|---|---|
| text | STRING | — |