Heads up: some links are affiliate links. If you sign up through one, we may earn a commission. It costs you nothing. Read the affiliate disclosure.

Trend · AI audio

ElevenLabs SFX v3: a whole scene's sound from one sentence

Updated Oct 31, 2026 · By Mishal Brands Editorial

Checked Oct 31, 2026. Everything below comes from ElevenLabs' own announcement on Oct 29, 2026, its developer documentation and its pricing page, all linked at the bottom. We have not generated audio with SFX v3 ourselves.

The short answer

SFX v3 is ElevenLabs' newest sound effects model, released on Oct 29, 2026. It is built to take one written description that names several sounds at once and get all of them into the audio. ElevenLabs' own example is a football clip that needs a crowd cheering, a ball dropping and the same crowd booing a few seconds later. Before this version, the company says, getting all of that from one prompt took trial and error. It is available now in the ElevenLabs web app, in the mobile app, and to developers through the API.

What actually changed

ElevenLabs lists three improvements over the previous version, and all three are the company's own benchmark results rather than an outside test.

Several sounds in one prompt. It follows a description that asks for more than one sound more closely, so more of what you wrote ends up in the result.
Music and vocals. It can carry a sung phrase or a musical line mixed in with everyday sound. ElevenLabs' documentation uses "Gospel choir singing hello" as an example.
Cleaner tones. It is less likely to crackle or clip on sounds that hold a note, such as gongs, chimes and whistles.

How to use it

In the browser. In ElevenLabs' creative app, sound effects are generated in the Sound Effects tab, in Flows, in Studio and in the Music Editor. The same model is in the mobile app.

In the API. It is the same sound effects endpoint as before. Set the model to eleven_text_to_sound_v3 and it uses the new version. If you change nothing, you get the older model, because v2 is still the default. Accounts without access get an Unknown model error rather than a silent fallback, so you will know.

There are two settings worth knowing. Duration can be set anywhere from 0.1 to 30 seconds; left alone, the model picks a length from your description. Prompt influence decides how literally it reads you. High follows the words closely; low leaves room for the model to add its own variation.

Prompts that work

ElevenLabs' own guidance is to describe the action, the setting, the atmosphere and the performance, instead of naming a broad category like "explosion". Its documentation gives a striking example of why wording matters: "a couple arguing in French" produced better results than "a couple fighting in French".

The blog suggests a simple loop: start with the main action, add the sounds around it, listen to how they sit together, then adjust the wording. Five starting points, built on that advice:

A busy coffee shop. "Espresso machine hissing, cups clinking, low chatter, rain on the front window."
A wet city street at night. "A car passing on wet asphalt, a distant siren, one dog barking twice, then quiet."
A football moment. "Crowd cheering, a ball dropping, the same crowd booing a few seconds later."
A children's party. "Balloons popping, a room of children singing happy birthday, a cheer at the end."
A room tone you can loop. "Soft rain on a window, quiet room, no voices." Then turn on looping and it repeats without a seam.

What it costs

Sound effects are paid for out of the same monthly credit pool as everything else on an ElevenLabs plan. ElevenLabs' pricing page lists a sound effect at 200 credits per generation, and 40 credits per second when you set the duration yourself.

Plan	Price and credits	What it means for sound effects
Free	$0, 10,000 credits a month	About 50 generations a month at 200 credits each. ElevenLabs lists the commercial licence as a paid feature, not a free one.
Starter	$6 a month, 30,000 credits	Adds the commercial licence, so you can use the audio in work you sell.
Creator	$22 a month, 121,000 credits	The plan ElevenLabs marks as most popular.
Pro	$99 a month, 600,000 credits	Adds higher-quality audio output through the API.
Scale and Business	$299 and $990 a month	Team seats and much larger credit pools.

Each generation is capped at 30 seconds. Output is MP3 for everything, with WAV at 48 kHz available for effects that do not loop. A one-minute scene means stitching a few generations together in the editor.

The catches

Access is limited for now. ElevenLabs says SFX v3 may not be available in every region, and accounts without access get an error rather than the old model.
The improvements are ElevenLabs' own claims. As FrameSignal noted on Oct 30, so far the only word on how much better it is comes from the company itself. ElevenLabs also recommends comparing v2 and v3 on your own prompts before you switch.
Exact counts are unreliable. ElevenLabs' documentation says a request for an exact number of repeats, such as a triangle struck four times, is not always followed.
Thirty seconds at a time. Long ambience beds have to be built from looped or joined clips.
Credits are shared. Every product on the plan draws from one pool, so heavy sound effects work leaves less for voiceover the same month.

Why it matters

Sound is the part of a short video that most people skip, because finding and layering stock effects is slow. If one sentence can produce a believable scene, the amount of work between a silent AI video clip and a finished one drops sharply. That is the part worth testing yourself: take a silent clip, write one line describing what the place should sound like, and see whether the first result is usable.

Related on Mishal Brands

Sources (checked Oct 31, 2026)

ElevenLabs, "Introducing SFX v3 in ElevenCreative and ElevenAPI" (Oct 29, 2026): https://elevenlabs.io/blog/sound-effects-v3
ElevenLabs developer docs, models: https://elevenlabs.io/docs/models
ElevenLabs developer docs, sound effects: https://elevenlabs.io/docs/capabilities/sound-effects
ElevenLabs pricing: https://elevenlabs.io/pricing
FrameSignal, "ElevenLabs says its new AI sound maker gets more of a busy scene's sounds from one written request" (Oct 30, 2026): https://framesignal.com/research/elevenlabs-sfx-v3-ai-sound-effects-whole-scene

Disclosure: some links on this page are affiliate links. If you sign up through one, we may earn a commission at no cost to you. No company paid for or reviewed this page. See our affiliate disclosure and how we review.