ElevenLabs Pro - Voice Settings Preset

ElevenLabs Pro - Voice Settings Preset is the pack's recipe card. Five voice-settings sliders in, one voice_settings JSON blob out - no API key, no network call, no credits. It's pure Python doing nothing more glamorous than json.dumps on five numbers, and it's genuinely useful for the same reason recipe cards are: it pins down the settings that make your narrator sound like your narrator.

The honest caveat: the TTS generation nodes in this pack build their own voice_settings internally from their own sliders, so this JSON doesn't plug into them as an input. What it does give you is a single, readable, reusable definition of your voice's behavior - keep the "documentary narrator" preset here and the "hype trailer" preset next to it, copy the JSON into a raw API call if you're doing anything outside ComfyUI, or just leave it in your workflow as self-documentation that says "these are the numbers this project uses." That's a small thing until your third workflow drifts to different settings and your narrator mysteriously changes character.

The settings it wraps

• stability (0–1, default 0.5) - lower is more expressive and emotional, higher is more consistent and monotone. The tooltip's shorthand is worth keeping: under 0.5 creative, at 0.5 natural, over 0.5 robust.
• similarity_boost (default 0.75) - how closely the AI sticks to the original voice. Too high and you invite artifacts.
• style (0–1, default 0) - exaggeration on top of the voice. The tooltip recommends 0 for most use cases, and that's a fair default; style can make things sound theatrical fast.
• speed (0.5–2, default 1.0) - straight playback speed multiplier.
• use_speaker_boost (default True) - boosts similarity to the source speaker at a small latency cost.

Install

ComfyUI Manager → search ComfyUI-ElevenLabs-Pro → Install → restart, or:

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

Restart ComfyUI. Dependencies are requests and soundfile (used by the pack's audio nodes), no model downloads - and because this node never touches the network, you don't even need an API key to use it.

Where people get burned

Honestly, people don't get burned by this node - the main disappointment is expecting it to feed its JSON into the TTS node and finding a string with no socket to plug into. Set that expectation, treat it as your canonical settings store, and it earns its place in a workflow quietly. It's the node you install the pack for last and reach for every project after.