Skip to content

How to use SFX v3: a guide to the ElevenLabs sound effects model

Sound designer in headphones types on a laptop as waveforms turn into a slamming door, shattering glass and an anvil Guides

SFX v3 is the third version of the model behind the ElevenLabs sound effects generator, released on October 29, 2026. To use it, open the Sound Effects tab in ElevenCreative, describe the sound and press generate. SFX v3 is already the default model there. In the API, send the model ID eleven_text_to_sound_v3, because the endpoint still defaults to v2.

This guide covers both routes, the key settings, prompts for several sounds at once, and prices. Facts were checked against ElevenLabs' docs and pricing on November 16.

What ElevenLabs SFX v3 does better than v2

ElevenLabs lists three improvements. SFX v3 follows prompts with several instructions more closely, so one generation can hold a whole scene. It handles music and vocals inside a sound effect better. It is also less likely to clip on sharp, high-frequency sounds such as chimes, gongs and whistles.

The company says v3 sounds cleaner than v2 in its own benchmark testing, but it has published no numbers. Our news story on the SFX v3 launch covers the announcement. The limits did not change: clips run up to 30 seconds and come out as MP3 or WAV. Here is how to try it.

How to use SFX v3 in ElevenCreative

The web app is the fastest way to start, and it needs no code.

  1. Sign in at elevenlabs.io and open Sound Effects under Playground in the sidebar.
  2. Type a description of the sound in the text box, up to 450 characters.
  3. Set the duration, or leave it on auto to let the model choose.
  4. Turn on Looping if you need a sound that repeats without a seam.
  5. Press the arrow. Each generation returns four versions of the sound.
  6. Open the History tab to play, star or download a result as MP3 or WAV.

SFX v3 is also the default model in Studio, the ElevenCreative editor, and in Flows. A menu in the Sound Effects tab still lets you switch back to v2 to compare. The three settings in steps 3 and 4 shape the result as much as the prompt does.

The three settings: duration, looping and prompt influence

Duration runs from 0.5 to 30 seconds. Auto works well for one-shot sounds like a door slam. Set it by hand for ambience and scenes, where the model needs room for every event.

Looping creates a clip whose end joins its start, for background beds such as rain or engine hum. For ambience, a longer duration of 20 to 30 seconds gives the loop room to settle.

Prompt influence ranges from 0 to 1, with a default of 0.3. Higher values follow your text more literally, and lower ones give more varied results. With good settings in place, the prompt does the rest.

SFX v3 prompt guide: how to describe a sound

ElevenLabs advises describing the action, setting, atmosphere and performance you want, rather than a broad category. "Footsteps on gravel approaching slowly" works better than "footsteps". A single sound needs a short, concrete prompt:

  • Glass shattering on concrete
  • Heavy wooden door creaking open
  • Thunder rumbling in the distance

For a scene with several sounds, list the events in the order they should happen. SFX v3 renders them together as one clip:

  • Football crowd cheering, a ball drop, then the crowd booing
  • Footsteps on gravel, then a metallic door opens
  • Rain against a window in a quiet apartment, a kettle starts to whistle

We ran three scene prompts from this guide through SFX v3 in ElevenCreative. Each generation returned four takes, and we kept one of each.

audio-thumbnail
SFX v3 – hammer on an anvil, ringing steel, then a steam hiss
0:00
/5
audio-thumbnail
SFX v3 – football crowd cheering, a ball drop, then booing
0:00
/8
audio-thumbnail
SFX v3 – rain against a window, a kettle starts to whistle
0:00
/25

Some guides report that beyond three or four sounds, elements start to drop out. For busy scenes, generate the bed and the key hits separately and layer them in Studio or an audio editor.

Sound design terms get recognizable results faster than piles of adjectives. The docs define the main ones: impact, whoosh, ambience, one-shot, loop, stem, braam, glitch and drone. For music elements, name the genre, tempo or key, as in "90s hip-hop drum loop, 90 BPM" or "vintage brass stabs in F minor".

Exact counts such as "triangle hit 4 times" are not always followed. Word choice also matters: ElevenLabs found "a couple arguing in French" worked better than "a couple fighting in French". Once a prompt works in the app, the same text works in the API.

How to use SFX v3 in the text to sound effects API

The ElevenLabs text to sound effects API takes a POST request to https://api.elevenlabs.io/v1/sound-generation. The model_id field selects the model, and it defaults to eleven_text_to_sound_v2. Name v3 explicitly:

curl -X POST "https://api.elevenlabs.io/v1/sound-generation?output_format=mp3_44100_128" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Hammer strikes on an anvil, ringing steel, then a steam hiss",
    "model_id": "eleven_text_to_sound_v3",
    "duration_seconds": 8,
    "prompt_influence": 0.5
  }' \
  --output forge.mp3

The API returns the audio file itself. In the official Python SDK, the same request looks like this:

import os
from elevenlabs.client import ElevenLabs

client = ElevenLabs(api_key=os.getenv("ELEVENLABS_API_KEY"))
audio = client.text_to_sound_effects.convert(
    text="Hammer strikes on an anvil, ringing steel, then a steam hiss",
    model_id="eleven_text_to_sound_v3",
    duration_seconds=8,
    prompt_influence=0.5,
)
with open("forge.mp3", "wb") as f:
    for chunk in audio:
        f.write(chunk)

The JavaScript SDK uses elevenlabs.textToSoundEffects.convert() with the same fields in camelCase, such as modelId. The output_format parameter accepts MP3, Opus, PCM, WAV and telephony formats. MP3 at 192 kbps needs the Creator plan, and 44.1 kHz WAV or PCM needs Pro. An "Unknown model" error means v3 is not yet available on the account.

Does looping work with SFX v3?

ElevenLabs' own pages disagree. The November 2 changelog lists "loop": true among the SFX v3 API options, and the models page lists looping as a v3 feature. The developer blog says looping "currently only works" with v2. Test a v3 loop in your player before you ship it. If you hear a click at the seam, generate the loop with eleven_text_to_sound_v2 until ElevenLabs settles the question. The last thing to check is the price.

SFX v3 pricing and commercial use

ElevenLabs bills sound effects in credits from your plan's shared pool. The pricing page lists 200 credits per generation, and the docs list 40 credits per second when you set a duration. On the API price list, sound effects cost $0.12 per minute.

  • Free: 10,000 credits, about 50 generations a month, MP3 downloads only, personal use only.
  • Starter: $6 a month, 30,000 credits, about 150 generations, commercial license and WAV.
  • Creator: $22 a month, 121,000 credits, about 605 generations.
  • Pro: $99 a month, 600,000 credits, about 3,000 generations.

The free plan does not allow commercial use. On any paid plan, ElevenLabs says you can use its sound effects royalty-free in commercial projects. Its terms say you keep the rights to your output. API users should plan the switch now: eleven_text_to_sound_v2 will be removed on March 1, 2027.

So is SFX v3 worth switching to?

For most people, yes, and in ElevenCreative the switch has already happened. SFX v3 is the better choice for scenes with several sounds, sounds with music or voices, and sharp sounds that clipped on v2. Write prompts as ordered events, set the duration for scenes, and keep it to three or four sounds per clip. In the API, add model_id: eleven_text_to_sound_v3 and compare your most common prompts against v2 first. Keep v2 for loops until the looping question is settled, and move everything else before March 2027.

FAQ

Short answers to common questions.

Is the ElevenLabs sound effects generator free?

Yes, within limits. The free plan includes Sound Effects with SFX v3 and 10,000 monthly credits, about 50 generations. Free results are for personal use only.

How long can an SFX v3 sound effect be?

Up to 30 seconds per generation, with a minimum of 0.5 seconds in the API. For longer ambience, generate a loop and repeat it in your editor.

On a paid plan, yes. ElevenLabs says paid users can use sound effects royalty-free in commercial projects, and you keep the rights to your output. Free plan output is for non-commercial use only.

How do I download ElevenLabs sound effects?

Open the History tab in Sound Effects and press download next to a result. Free plans get MP3, and paid plans can also choose 48 kHz WAV.

Can I add SFX v3 sound effects to a video?

Yes. Studio in ElevenCreative puts generated effects on a video timeline, and its Video to Music tool now adds sound effects with SFX v3.

Sources

Portrait of Felix Laurent, author at thehype.news

Felix Laurent

News Writer at thehype. Felix is a journalist with experience in marketing. He covers artificial intelligence and emerging technologies, focusing on the ideas, innovations, and trends shaping the future of tech.

More articles by Felix Laurent →
On air · radio.thehype.news ↗ AI news radio — 24/7