ElevenLabs already presents AI-generated variations of human voices and music. Now, it’ll let individuals create sound results for podcasts, motion pictures, or video games, too. The brand new Sound Results software can generate as much as 22 seconds of sounds based mostly on person prompts that may be mixed with the corporate’s voice and music platform, and it offers customers not less than 4 downloadable audio clip choices.
The corporate says it labored with the inventory media platform Shutterstock to construct a library and prepare its mannequin on its audio clips. Shutterstock has licensed its content material libraries to many AI firms, together with OpenAI, Meta, and Google.
Sound Results is free to make use of, however paid tiers can use the generated audio clips with industrial licenses, whereas free customers “should attribute ElevenLabs by together with ‘elevenlabs.io’ within the title.” ElevenLabs customers have a set character depend restrict when writing prompts, with free customers getting 10,000 characters per 30 days. For Sound Results, ElevenLabs says on its FAQs web page that it’ll take 40 characters per second from the allotment if customers set the audio clip length themselves. If utilizing the default audio length, every immediate request might be charged 200 characters.
Libraries with sound impact clips exist already available in the market for creators, filmmakers, and online game builders. However generally, these may be costly or have bother surfacing simply the precise kind of sound. ElevenLabs says in its weblog publish that it designed Sound Results “to generate wealthy and immersive soundscapes rapidly, affordably and at scale.”
Different AI builders are additionally creating their very own text-to-sound mills. Stability AI launched Steady Audio final yr, which creates audio clips of music and sound results, and Meta’s AudioCraft fashions generate pure sound (assume background noises like wind or visitors).