ElevenLabs - Fetch Voices

The AIS_EL_VoiceSelector dropdown is a static, baked-in menu of premade voices. This node is the opposite: it calls ElevenLabs and pulls back a live list of every voice your account can actually use - premade, cloned, generated, and professional. If you've cloned a voice or designed one, it shows up here. That's the entire difference between "the fixed catalog" and "what I own."

You give it an api_key, and it hits the voices endpoint. Three controls matter:

• search - filter the results by name, description, accent, or use case. Handy when you have a hundred voices and want only the ones tagged for audiobook work.
• category - narrow to premade, cloned, generated, or professional. Good for programmatically picking "give me all my clones" without parsing names.
• force_refresh - the pack caches voice lists to avoid hammering the API every queue (the node runs fresh each time thanks to IS_CHANGED, but the underlying voice list is cached). Flip this on when you've just created a new voice and want it to appear immediately.

Two outputs:

• voices_json - the full API response as a string, with every field (IDs, names, descriptions, preview URLs).
• voice_id_list - just the IDs, as a string. Useful to pipe into an LLM node or a selector so you can pick a voice dynamically instead of by hand.

How to actually use it

The natural loop is: clone or design a voice → run FetchVoices with force_refresh → grab its voice_id → wire it into TTS. In a more automated setup you can feed voice_id_list into a text-processing node and let a prompt or an LLM choose which voice speaks each line - that's how you'd script a multi-voice project without hand-entering a single ID. ElevenLabs returns roughly a preview URL per voice too, so if you want to audition them visually in a browser-preview node, the JSON has everything you need.

Installing it

One of fifteen ElevenLabs nodes in the ComfyUI API Toolkit pack. Manager: search "API Toolkit". Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/IxMxAMAR/ComfyUI-API-Toolkit
cd ComfyUI-API-Toolkit
pip install -r requirements.txt

Restart ComfyUI. Requires requests and soundfile from requirements.txt; missing deps silently disable only the affected service.

Gotchas

• This is a real API call that costs you nothing in characters but does count against rate limits - which is why the cache exists. If you're looping, lean on the cache and only force_refresh when you genuinely created a voice.
• The voice_id_list output is a string, not a true ComfyUI list, so downstream nodes that expect a Python list won't take it directly. Treat it as text to parse.
• A 401 means the key's wrong; the node's shared retry handles transient 429s/5xx for you, but not a bad key.