ElevenLabs AI Sound Effects: Prompts and API Options (2026)

ElevenLabs AI sound effects is the text-to-sound feature inside ElevenLabs: you describe a sound in plain English, such as “a heavy wooden door creaking open in a stone hallway”, and it returns short audio clips that match the description. ElevenLabs built its name on realistic speech, and sound effects take the same idea from voices to footsteps, impacts, ambiences and UI sounds. This guide covers how the feature works, how to write prompts that give you usable clips on the first try, what the licensing covers, how to get clips into a project, and how to generate sound and ElevenLabs voices from code with one API key.

What are ElevenLabs AI sound effects?

A sound effect is any artificially created or enhanced sound used to support a piece of media: a punch in a film, a notification chime in an app, rain behind a podcast intro. Traditionally you got them from a recorded library or a foley artist (someone who performs sounds live in sync with picture). ElevenLabs AI sound effects replace the search step with generation. You write a prompt, the model synthesises a few variations, and you keep the one that fits.

In practice, the tool gives you:

• A prompt box. You describe the sound you want. Detail matters, and more on that below.
• Duration control. You can let the model choose a length or set one yourself.
• Prompt influence. A setting that controls how literally the model follows your words compared with how much it improvises.
• Several variations per prompt. You audition them and download the best one.
• A browsable library. ElevenLabs also hosts downloadable effects grouped by category (devices, nature, vehicles and so on). It’s useful when someone has already generated what you need.

The feature sits next to the products most people know ElevenLabs for: the ElevenLabs text to speech engine, the ElevenLabs AI voice generator with its large voice library, and voice cloning. That’s why so many people search for ElevenLabs AI sound effects: they’re already making voiceovers in ElevenLabs and want the sound design from the same place.

The model can’t read your timeline. It doesn’t know your video, your music or your mix. It returns isolated clips, and placing them, timing them and balancing their levels is still your job. If you keep that limit in mind, you’ll judge the tool fairly and use it well.

How to write prompts for AI sound effects

Most disappointing results come from vague prompts. “Explosion” gives you a generic boom. “Distant artillery explosion echoing across an open valley, low rumble tail” gives you something you can actually place in a scene. A reliable structure has four parts:

1. Source: what makes the sound (a ceramic mug, a diesel truck, a small dog).
2. Action: what it does (set down on a wooden table, idling then pulling away, barking twice).
3. Space: where it happens (small tiled bathroom, open field, inside a car).
4. Texture and timing: how it should feel (sharp, muffled, metallic, slow fade, single hit).

Some examples that work well with ElevenLabs AI sound effects and with sound models in general:

Weak prompt	Stronger prompt
Footsteps	Slow footsteps in leather boots on a creaky wooden floor, indoor, close microphone
Rain	Steady light rain on a car roof, heard from inside the car, no thunder
Whoosh	Fast airy whoosh for a UI transition, clean, no low end, under one second
Crowd	Small cafe ambience, quiet murmured conversation, occasional cup clinks, no music
Sci-fi sound	Short futuristic door sliding open with a pneumatic hiss and soft electronic chirp

Tips that save regenerations

• Say what you don’t want. “No music”, “no voices” and “no reverb” all help, because models like to fill space.
• Ask for one event per clip. If you want a door creak and a slam, generate them separately and line them up in your editor. You get control over timing and can reuse each piece.
• Use audio vocabulary. Words like “close-miked”, “distant”, “dry”, “roomy”, “tail” and “transient” (the sharp start of a sound) steer the result.
• Raise prompt influence for literal sounds. Keep it lower when you’re exploring stylised or abstract textures.

Are ElevenLabs AI sound effects free, and can you use them commercially?

ElevenLabs runs a free plan with a small monthly credit allowance, and sound generations draw from the same pool as speech. That’s enough to learn how prompts behave, but a real sound pass on a short video can use up a free allowance quickly. You’ll usually generate several takes per sound, and only one survives.

On rights, ElevenLabs states that its sound effects are royalty free and can be used in commercial projects. Two practical caveats:

• Plan terms still matter. Free tiers at AI audio companies often carry attribution requirements or narrower commercial rights than paid tiers. Check the terms that applied to the plan you were on when you generated the clip, not the plan you’re on today.
• Keep records. Save the prompt, the date and the file for anything you ship to a client. If anyone asks where a sound came from, you can answer in a minute.

Also be aware that generated audio can sometimes sound close to recognisable material. That’s rarely a problem for footsteps and wind. It can matter more for anything resembling a branded sound, like a famous startup chime. Don’t prompt for trademarked sounds, and you avoid the issue.

How to put ElevenLabs sound effects into a project

A common question from ElevenLabs users is how to get a generated effect into a Studio project (the long-form editor formerly called Projects). The interface moves around between releases, but the workflow below holds up whatever the current layout is:

1. Generate and download. Audition the variations from ElevenLabs AI sound effects and download the keeper as an audio file. Name it clearly, for example door_creak_stone_01.mp3.
2. Import it as its own track. In Studio, or in whatever video editor you finish in, place the effect on a separate track from voice and music. Keep at least three lanes: dialogue, effects and music.
3. Time it to picture or speech. Line up the transient with the on-screen action. A few frames early often feels better than exactly on the frame.
4. Balance it. Effects should support speech. Duck (lower) the effect or music under narration, and check the mix on headphones and on a phone speaker.
5. Export a stem. If a client may request changes, export the effects track on its own as well as the full mix.

For long ambiences, generate a short clean segment and loop it in the editor instead of asking for one long clip. Long single generations tend to drift in character, and drift is harder to fix than a loop seam you can hide with a short crossfade.

Generating sound and ElevenLabs voices through one API

The browser tool is fine for one video. Once you’re building something, such as a game that needs dozens of UI sounds, a pipeline that produces narrated explainers, or an app that generates audio on request, you want an API. ElevenLabs offers its own developer API. If your product also needs images, video or chat, managing a separate account and bill for each provider gets tedious fast.

Unicorn API gives you one API key and one bill for image, video, audio and chat models, priced in credits. For audio, the catalog includes these models:

• Elevenlabs Text to Speech Turbo V2.5: ElevenLabs text to speech at 6 credits per 1,000 characters.
• Elevenlabs Text to Speech Multilingual V2: 12 credits per 1,000 characters, with a language_code field for multilingual work.
• Elevenlabs Text to Dialogue V3: multi-speaker dialogue at 14 credits per request.
• Ai Music API Sounds: generated sounds from a text prompt at 2.5 credits per track, with optional sound_loop, sound_tempo and sound_key fields.

The ElevenLabs sound effects model itself isn’t in the Unicorn API catalog. What you get is ElevenLabs voices plus a separate sound-generation model under the same key. For many projects that’s exactly the pairing ElevenLabs AI sound effects users want: ElevenLabs narration with generated sound underneath. The full list is in the model catalog.

The request flow

Every media model on Unicorn API uses the same asynchronous pattern. Asynchronous means you don’t wait on one long HTTP call:

1. Optionally ask POST /v1/models/price for the exact cost of the request.
2. Submit the job with POST /v1/jobs.
3. Poll GET /v1/jobs/{id} until status is "succeeded".
4. Download the file from job.outputs[0].url.

Python example: one sound effect plus a narration line

Create a key on the API keys page and export it as UNICORN_API_KEY. The script below generates a sound with Ai Music API Sounds and a narration line with ElevenLabs Turbo V2.5, then downloads both.

import os
import time
import requests

BASE_URL = "https://api.unicornapi.net"
HEADERS = {
    "Authorization": f"Bearer {os.environ['UNICORN_API_KEY']}",
    "Content-Type": "application/json",
}


def price(body):
    r = requests.post(f"{BASE_URL}/v1/models/price", json=body, headers=HEADERS)
    r.raise_for_status()
    return r.json()


def run_job(body, timeout=300):
    r = requests.post(f"{BASE_URL}/v1/jobs", json=body, headers=HEADERS)
    r.raise_for_status()
    job = r.json()

    deadline = time.time() + timeout
    while time.time() < deadline:
        r = requests.get(f"{BASE_URL}/v1/jobs/{job['id']}", headers=HEADERS)
        r.raise_for_status()
        job = r.json()
        if job["status"] == "succeeded":
            return job["outputs"][0]["url"]
        if job["status"] in ("failed", "canceled"):
            raise RuntimeError(f"Job {job['id']} ended with status {job['status']}")
        time.sleep(3)
    raise TimeoutError(f"Job {job['id']} did not finish in {timeout}s")


def download(url, path):
    r = requests.get(url)
    r.raise_for_status()
    with open(path, "wb") as f:
        f.write(r.content)


sound_body = {
    "model": "ai-music-api-sounds",
    "input": {
        "model": "V5",
        "prompt": "Steady light rain on a car roof, heard from inside the car, no thunder, no music",
    },
}

voice_body = {
    "model": "elevenlabs-text-to-speech-turbo-2-5",
    "input": {
        "text": "The storm had been following us for an hour, and nobody wanted to say it out loud.",
        "voice": "EkK5I93UQWFDigLMpZcX",
    },
}

print("Sound price:", price(sound_body))
print("Voice price:", price(voice_body))

download(run_job(sound_body), "rain_car_roof.mp3")
download(run_job(voice_body), "narration_01.mp3")
print("Saved rain_car_roof.mp3 and narration_01.mp3")

The same request as a quick curl call, if you just want to see a job come back:

curl -X POST https://api.unicornapi.net/v1/jobs \
  -H "Authorization: Bearer $UNICORN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"ai-music-api-sounds","input":{"model":"V5","prompt":"Fast airy whoosh for a UI transition, clean, no low end"}}'

For two characters talking, swap in Text to Dialogue V3 and pass a dialogue array of {"text": ..., "voice": ...} entries. Field details for every model are in the API documentation. If you’d rather test prompts before writing code, the audio playground runs the same models in the browser. If you work inside Cursor or Claude Code, the agents page explains how to call the API through MCP.

What does a sound and voice pass cost? A worked example

Here’s the cost of a short narrated explainer: eight sound effects and about 2,400 characters of narration (roughly a two-minute script). The costs below use only the per-request prices listed above.

Item	Model	Pricing rule	Credits
8 sound effects, one request each	Ai Music API Sounds	2.5 credits per track, rounded up to whole credits, so 3 per single-track request	24
Narration, 2,400 characters	ElevenLabs Turbo V2.5	6 credits × ceil(2,400 / 1,000) = 6 × 3	18
Total			42

On a monthly plan, credits cost $0.01 each, so this pass costs $0.42. Plans are Starter at $25, Creator at $50, Studio at $100 and Pro at $200 per month. With top-up credits at $0.015 each, which never expire, the same pass costs $0.63. If you use Multilingual V2 for the narration, the narration line becomes 12 × 3 = 36 credits. Regenerations are the real cost driver: if you average three takes per effect, the sound line triples. That’s why writing a good prompt first is the cheapest optimisation you have. Plan details are on the pricing page.

Two billing details to design around:

• Text to speech bills in 1,000-character blocks. A 1,001-character request costs the same as a 2,000-character one. Split long scripts at natural paragraph breaks close to multiples of 1,000 characters, not at random points.
• Rounding applies per request. Use the price endpoint before batch runs so your estimate matches the bill exactly.

Common mistakes with AI sound effects

The same handful of problems come up whether people use ElevenLabs AI sound effects in the browser or a sound model through an API:

• One prompt for a whole scene. “Busy kitchen with a phone ringing and someone dropping a plate” gives you a muddle. Generate the ambience, the ring and the smash separately, then layer them.
• Keeping the first take. Generation is cheap compared with your time in the edit. Audition several takes and pick deliberately.
• Ignoring loudness. Generated clips arrive at different levels. Normalise them before you mix, or your effects will jump out unevenly.
• Effects masking speech. If the narration is the point, the effects sit underneath it. Cut low-frequency rumble under the voice and lower the effects during lines.
• Not saving prompts. A prompt that produced a perfect whoosh is an asset. Keep a prompt sheet next to your project files so you can reproduce or extend a sound set later.
• No timeout in code. Polling loops without a deadline hang forever when a job fails. The example above handles failed jobs and timeouts for that reason.

ElevenLabs app or an API: which fits your work?

Both routes are reasonable, and many people use both: the browser to explore, the API to produce at scale.

Need	Better fit	Why
A handful of effects for one video	ElevenLabs web app	Fast auditioning of variations and direct download
Browsing ready-made sounds	ElevenLabs sound library	No generation needed if someone already made it
Automated narration plus sound in a product	Unicorn API	ElevenLabs voices and a sound model behind one key and one credit balance
Audio alongside images, video or chat	Unicorn API	One bill across media types instead of one account per provider
Loops with set tempo or key	Ai Music API Sounds	Exposes sound_loop, sound_tempo and sound_key inputs

If you’re comparing voices rather than effects, our guide to the best AI voice generator goes through the speech models in more depth. For a wider look at sound generation beyond ElevenLabs AI sound effects, see the AI sound effects generator guide. For full music tracks rather than short sounds, making a Suno AI song covers prompts and code for the music models, including Ai Music API Mashup.

Start with clear four-part prompts, keep one event per clip, mix effects under speech, and check prices before batch runs. That covers most of what separates usable ElevenLabs AI sound effects from clips that end up unused.