[DigitalToday reporter Chi-gyu Hwang] ElevenLabs, a startup specialising in artificial intelligence (AI) audio research and development, said on July 23 it updated its AI music generation service, ElevenMusic.

Through the update, ElevenLabs added References, which lets users specify a song’s style and mood using reference audio, and Vocals, which generates original songs using a user’s voice or voices from a library.

The Styles feature, which helps keep a music track’s distinctive style consistent, now officially supports the latest AI model, Music v2.

References is the official version of the reference-audio feature previewed when Music v2 was announced in June. If users upload reference audio lasting from 10 seconds to 5 minutes, Music v2 generates music to match the track’s style and mood. It allows subtle sound nuances that were difficult to express with text prompts alone to be delivered directly to the AI model using real sound.

Vocals is a new music production method that starts from a voice. Users can generate a completely original song sung in that voice simply by selecting their own voice or a vocal library. It can also be used by people without music production experience.

ElevenLabs said artist rights protection remained its top priority in this update. It stressed that all voices and songs uploaded to the platform are designed to be usable only after passing strict copyright checks before the generation stage.