Load Artists CSV

Load Artists CSV is the one node in this pack with real weight behind it. Its artists.csv ships with over three hundred entries - Ai Weiwei, Adolf Wölfli, Agnes Denes, all the way down - and turns them into a single dropdown that outputs the artist's name as a prompt tag. If you've ever spent a session flipping between "Gustav Klimt" and "Alphonse Mucha" and retyping the name into your prompt box, this is the fix. It's the node that quietly does for your style list what the pack does for everything else: keeps it in a file you can actually edit instead of in a pile of orphaned text nodes.

This is also the pack's most direct A1111-bridge node, and that's by design - the author built the whole pack around the style.csv idea that A1111 users already knew, the file where you saved your named prompt presets. Coming from A1111, you already have this muscle memory: name a preset, recall it later, batch across it.

How it works

Same engine as the rest of the pack. On load it reads custom_nodes/ComfyUI-CSV-Loader/CSV/artists.csv, skips the header, and parses name, positive, negative per line into a lookup dict. The positive prompt for an artist row is just the artist's name with a trailing comma - so the node's real job is "remember which artists I care about and give me the tag." The negative column is empty across the shipped file.

One honest caveat: artist-name prompting is checkpoint-dependent. SDXL-lineage models trained on booru-style data respond strongly to artist tags; the LLM-encoded 2026 checkpoints read them as words in an instruction and may largely ignore them. If you're on Flux or a Qwen-encoded model, this node still organizes your list, but the results will be blander than on an SDXL checkpoint.

The inputs and outputs that matter

• artists - the dropdown. Three hundred-ish entries; start typing and it filters, which is the whole point.
• positive prompt (STRING) - the artist tag. Into the positive CLIPTextEncode.
• negative prompt (STRING) - empty in the shipped file.

Since the output is plain text, you can do useful things with it that a hardcoded style list can't: concatenate it with other prompt chunks in your workflow, or feed it into a batch loop and generate the same subject across a dozen artists to compare styles. That's the workflow this node is genuinely good for.

Installing it

Same story as the whole pack - no dependencies, no models, no pip step:

cd ComfyUI/custom_nodes
git clone https://github.com/PCMonsterx/ComfyUI-CSV-Loader

...or unzip it there, or install via ComfyUI Manager by searching "ComfyUI-CSV-Loader", then restart. The folder name must stay ComfyUI-CSV-Loader because the loader's path is hardcoded to it.

Common issues

• Dropdown shows "Error loading artists.csv, check the console". The file isn't at the hardcoded path - usually because the folder got renamed during extraction, or you're on Linux/macOS where the Windows-style path in the source won't resolve. Windows is the platform this pack was built and tested on.
• A line with too few columns fails the whole file. The parser builds everything in one dict comprehension, so one stray row poisons the load. Keep rows to three columns, and quote any value containing a comma.
• I added artists and they're not in the dropdown. Refresh ComfyUI; the list is built when the node loads.

The maintenance caveat applies here most of all: the pack is unmaintained and the README points you at Dr.Lata nodes and forks. For the artists list specifically that's almost a non-issue, because the value isn't the shipped list - it's the file. Delete the entries you'll never use, keep your ten working artists, and this node becomes a permanent part of your style toolkit.