ElevenLabs Sound Effects Generator (SFX v3)
Generate sound effects with ElevenLabs SFX v3, the newest ElevenLabs text to sound model. Describe a single sound or a full scene, choose an exact length or let the model decide, and loop it for background audio.
Examples
A car engine starts, revs twice, tires screech around a corner, then a loud crash with broken glass falling
Name the source, the action and the space. Several sounds in a row work too: describe them in order.
The model picks a length that fits the sound.
The end flows back into the start, for background ambience and game loops.
Auto length reserves 12 credits; anything a shorter sound does not use is returned when it is ready
Examples
A car engine starts, revs twice, tires screech around a corner, then a loud crash with broken glass falling
Spacious braam suitable for high-impact movie trailer moments
Campfire crackling and popping at night with crickets in the background
Footsteps of one person walking slowly on crunchy gravel, close microphone
Massive dragon roar echoing across a mountain valley, followed by heavy wing flaps
Short 8-bit chiptune victory jingle for completing a level
Every example was generated from the prompt shown, with the length and loop setting listed. Volume was evened out for this page.
What's new in SFX v3
ElevenLabs released SFX v3 on October 29, 2026 as the successor to SFX v2. Three things improved.
Multi-sound prompts
SFX v3 follows several instructions in one prompt more reliably, so more of the sounds you list end up in the clip. Describe a scene in order, like a crowd cheering, a ball dropping and the crowd booing, and hear it as one take.
Music and vocal elements
Prompts that mix in a musical phrase or a vocal performance, such as a chant, a hum or a short jingle, come out more convincingly than with SFX v2.
Cleaner, more natural audio
Fewer harsh or crackly artifacts on tonal sounds like gongs, chimes and whistles, which keep their ring and resonance.
SFX v3 parameters on Veevid
Every control the model offers, and how it is set here.
- Model
- ElevenLabs SFX v3 (API model ID eleven_text_to_sound_v3)
- Prompt
- Up to 450 characters describing the sound or scene
- Duration
- 0.5 to 30 seconds in 0.5-second steps, or auto (the model picks the length)
- Loop
- On or off. On makes the end connect back to the start
- Prompt influence
- 0 to 1, default 0.3. Higher follows your words more literally with less variation
- Output
- MP3, 44.1 kHz stereo, 128 kbps
- Price
- 0.4 credits per second, rounded up, minimum 1 credit. 10 seconds = 4 credits
How to use ElevenLabs SFX v3
The same controls the model exposes, in three steps.
Write the prompt
Describe the sound, or list several sounds in the order they should happen. Up to 450 characters.
Tune duration, loop and influence
Pick 0.5 to 30 seconds or auto, switch on loop for repeating ambience, and raise prompt influence under Advanced if results drift from your wording.
Generate, compare, download
Generate a few takes, keep the one you like and download it as MP3. Each take is saved in My Creations.
ElevenLabs SFX v3 FAQ
About the model and how it runs on Veevid.
SFX v3 is the third generation of the ElevenLabs sound effects model. It turns a text description into a sound effect of up to 30 seconds and was released on October 29, 2026.
ElevenLabs describes v3 as better at prompts with several sounds, more convincing with music and vocals, and cleaner on tonal sounds like gongs and chimes. The controls are the same: a text prompt, duration, loop and prompt influence.
No. SFX v3 runs on Veevid and is paid with Veevid credits, so a Veevid account is all you need.
Yes. With loop turned on, the generated sound is designed to repeat without a gap. In our own tests a 10-second loop had no fade at either end and the join was smooth.
It sets how literally the model follows your words, from 0 to 1. The default of 0.3 leaves room for variation; around 0.6 to 0.8 sticks closer to the prompt and makes repeated takes more alike.
0.4 credits per second of audio, rounded up, with a minimum of 1 credit, so 10 seconds costs 4 credits and the 30-second maximum costs 12. With auto duration, 12 credits are reserved and the unused part is returned as soon as the sound is ready.
It can produce vocal elements like a chanting crowd, but it is not a voice model. For spoken lines use Eleven v4 text to speech.
More audio tools
Pair sound effects with voice, music and clean audio.
Try ElevenLabs SFX v3
Write a prompt, set the duration and hear the result in seconds.
Generate with SFX v3