Cheers, a ball and boos in one prompt: ElevenLabs launches sound-effects AI SFX v3
The new model creates several sounds at once from a single text description and, the company says, renders music and vocals more convincingly and sounds such as gongs more cleanly.
Photo: Rafał Masłow / Wikimedia Commons, CC BY-SA 4.0 (cropped and resized)
Video: YouTube · ElevenLabs · youtube.com
AI audio specialist ElevenLabs on Thursday unveiled SFX v3, the third version of its model that turns short text prompts into sound effects. According to the company, the model follows several instructions within a single prompt more reliably, so more of the described sounds actually end up in the finished clip. The official ElevenLabs video shows examples.
ElevenLabs gives a football scene as an example: a cheering crowd, a ball dropping and, a few seconds later, booing – all from one prompt. The model is also said to be better when a melody or a vocal performance has to be combined with ambient sounds. In addition, the audio sounds cleaner and more natural, with fewer artefacts on gongs, chimes and whistles. These claims are based on the company's own comparison tests against SFX v2; there are no independent tests.
SFX v3 is available in the ElevenCreative platform – including Studio and the Music Editor – via the ElevenAPI developer interface and in the ElevenLabs mobile app. The announcement gives no details on pricing or maximum clip length.