AI4ArtsEd Random Language Selector
Twelve random languages, from Yoruba to Norwegian, for prompt translation chains
- language_1
- language_2
- language_3
- language_4
- language_5
- language_6
- language_7
- language_8
- language_9
- language_10
- language_11
- language_12
AI4ArtsEd Random Language Selector does one thing: samples twelve languages from a curated list of about fifty and hands them to you as twelve STRING outputs. No inputs, no LLM, no network. Run it and you get language_1 through language_12 - a mix that might be Mandarin, Swahili, Ukrainian, Yoruba, Māori-adjacent territory, and Norwegian all in one shot.
The interesting bit is which languages made the list. The pack's docs describe it as "a curated list including both major and marginalized world languages," and that's true: English, Spanish, and Hindi sit next to Hausa, Igbo, Kinyarwanda, Xhosa, Quechua, and others that most prompt-translation workflows simply ignore. The stated use case in the README is stille_post-style translation chains - workflows where you translate a prompt into another language and back, or where you want the model to absorb the semantics of a prompt written in a language it wasn't prompted in. It's a prompt-variety and cultural-education tool more than a utility.
How it works
The implementation is transparent: a hardcoded list of ~50 language names, and random.sample(..., 12) picks twelve distinct ones on each execution. Because there's no IS_CHANGED and no inputs, the node runs when anything upstream changes and ComfyUI re-executes it - but since it has no inputs, a fresh draw needs something in the graph to change first. There's no seed to pin a specific draw.
Inputs and outputs:
- Inputs: none.
- Outputs:
language_1throughlanguage_12(twelve STRINGs). No guarantee of a particular language; all are distinct within a run.
Wiring it in
The outputs are meant to feed translation-oriented LLM nodes - ai4artsed_prompt_interception with a "translate into {language}" style prompt, or an external translate node. With twelve outputs and typically one translation stage, you'll usually wire one or two and ignore the rest, or feed them into ai4artsed_text_remix in random mode and let each run pick a language.
Installing it
Same as the whole pack:
cd ComfyUI/custom_nodes
git clone https://github.com/joeriben/ai4artsed_comfyui_nodes
# restart ComfyUI
Or search "AI4ArtsEd" in ComfyUI Manager. Zero dependencies beyond random.
Common issues
A language you wanted never shows up. With 12 drawn from ~50, any given run is a sample, not a superset. If you specifically want Guarani or Thai, this node won't guarantee it - it's random by design.
Reproducibility. There's no seed, so you can't replay a specific draw. For research this is mildly annoying; for prompting it's usually fine because you don't need the same twelve languages twice.
It's a tiny node with a specific mission - injecting linguistic variety into translation chains - and it does that without fuss. Just don't expect it to do anything with the languages. It picks them; you supply the translating.
Inputs (0)
No inputs
Outputs (12)
| Name | Type | Description |
|---|---|---|
| language_1 | STRING | — |
| language_2 | STRING | — |
| language_3 | STRING | — |
| language_4 | STRING | — |
| language_5 | STRING | — |
| language_6 | STRING | — |
| language_7 | STRING | — |
| language_8 | STRING | — |
| language_9 | STRING | — |
| language_10 | STRING | — |
| language_11 | STRING | — |
| language_12 | STRING | — |