The new ElevenLabs sound effects model better follows instructions at several events and produces a cleaner sound, on ElevenCreative, the mobile app and the API.
On October 29, ElevenLabs presented SFX v3, the third generation of its model that generates sound effects from a text description. The main advance claimed concerns the instructions describing several successive sounds: the company cites the example of a football sequence where the crowd cheers, the ball falls, then whistles fly, each element now arriving in its place.
The model also handles music and vocals better, including mixed with environmental noise, and produces less crackling and metallic artifacts on difficult sounds such as gongs, chimes or whistles. These gains are based on internal testing by ElevenLabs, which has not published any figures or methodology.
SFX v3 is available now in ElevenCreative (Flows, Sound Effects tab, Studio and music editor), in the mobile app and, for developers, via ElevenAPI.
Read the full article on the original source
Go to the source →