ElevenLabs Sound Effects
Foley on demand from a text prompt
- AUDIO
Type "thunder rumbling" or "door creaking in an empty hallway" and this node hands you a ready-to-use sound effect, up to 22 seconds of it. If you've ever paused a ComfyUI video render and thought "this looks great but sounds like nothing," this is the node that fixes that - it's the on-demand foley desk for your audio layer.
How it works
It's a text-to-audio call to ElevenLabs' /v1/sound-generation endpoint. Your text prompt, a duration_seconds, and a prompt_influence value go up as JSON, and the returned audio comes back as a standard ComfyUI AUDIO tensor. Clean and synchronous - no project IDs, no polling, no dashboard trips. This one behaves the way you'd hope: run it, get audio.
Inputs that matter
text- the description of the sound. This is a prompt, same as the image side: "dog barking in the distance" (the default) beats "dog." Include distance, room, mood, source.duration_seconds- 0.5 to 22 seconds, default 5. Shots that need a loop or a pad want the top end; a single hit wants the bottom.prompt_influence- 0 to 1, default 0.3. This is the interesting one: how literally the model follows your prompt versus improvising. Low values give you more creative latitude (and can ignore bad prompts), high values pin it to what you wrote. If your prompt is precise, push it up; if you're exploring, leave it low.input_text(optional) - a force-input variant oftext, for when you want the prompt wired in from elsewhere in the graph instead of typed.
Output: one AUDIO socket.
Install
Same pack, same commands:
cd ComfyUI/custom_nodes
git clone https://github.com/karthikg-09/Elevenlabs-ComfyUI.git
cd Elevenlabs-ComfyUI && pip install -r requirements.txt
or ComfyUI Manager → search "Elevenlabs-ComfyUI" → restart. Under Add Node → ElevenLabs.
The honest take
This sits in a weird spot relative to the open ecosystem. For generative foley you have real local options - MMAudio is the community workhorse for matching sound to video, and Stable Audio is a solid open-weights sound-design tool - so ElevenLabs isn't the only door anymore. What this node buys you is convenience: no model files, no dependency conflicts, one prompt away from a usable effect. For quick placeholder foley on a consumer card it's genuinely handy, and since effects are cheap short calls, it won't eat your credit balance the way TTS volume can.
Where it stumbles is length and control: 22 seconds is plenty for an impact or a whoosh but useless for continuous ambience, and prompt_influence is a blunt instrument next to something like Stable Audio's LoRA tuning. Reach for this when you need a single good effect in a couple of minutes. Reach for the local tools when foley becomes a real part of your pipeline. Errors, as everywhere in this pack, show up in the ComfyUI console with the HTTP status and the API's actual response text.
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| api_key | STRING | — | |
| text | STRING | Dog barking in the distance | — |
| duration_seconds | FLOAT | 5.00.5–22 | — |
| prompt_influence | FLOAT | 0.300–1 | — |
| input_textopt | STRING | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| AUDIO | AUDIO | — |