ElevenLabs Text to Sound Effects

Every video project hits the same wall: you need a sound effect and you don't have one. ElevenLabsTextToSoundEffects is the wall-remover. Type "thunder rolling across a valley," and a few seconds later you've got an AUDIO clip of exactly that, generated from nothing but text. It's the audio foley kitchen of the ElevenLabs group, and for a one-off effect it's dramatically faster than scrubbing a stock library.

Partner-node mechanics: the prompt goes to ElevenLabs' hosted eleven_sfx_v2 model through Comfy's API proxy, billed per minute of generated audio from your Comfy account credits (Settings → Credits). No ElevenLabs key, ships with ComfyUI core, nothing to install.

The inputs that matter

• text - the description. Be concrete and sensory: "wooden ship creaking in rough water" beats "boat sounds."
• model - eleven_sfx_v2, which reveals three sub-parameters worth knowing:
  • duration - 0.5 to 30 seconds of output. Most effects want a short clip; pick the length and pay only for that.
  • loop - toggle to make a smoothly looping effect. This is the one people discover late; it's the difference between a one-shot boom and an endless ambient hum.
  • prompt_influence - 0 to 1, default 0.3. How tightly the output follows your text. Crank it up for a literal match, drop it for something looser and more creative.
• output_format - mp3 or opus, default fine.

Output and wiring

Single AUDIO output into Preview Audio, Save Audio (Advanced), or straight into a concat with dialogue or music. Because it's a real audio signal, you can layer it, equalize it, or duck it under narration - no different from any other clip.

Where people get burned

• Default duration is 5 seconds, and some effects (a door creak, a footstep) are a single instant - you'll get a padded tail. Set duration to match the effect, not the field's default.
• prompt_influence is low by default (0.3). If your effect comes back loose and interpretive, that's the knob; push it toward 1.0 for literal adherence.
• Per-minute billing on a 30-second max is cheap, but iterate on a 2-second test before locking the full-length version.
• It generates effects, not music beds. For a composed track you want the text-to-music nodes (SoniloTextToMusic), not this.

For ambiance, transitions, and one-off Foley, it's the fastest text-to-sound path in the graph. Type it, hear it, save it.