YouTube is rolling out an experimental AI tool that allows select content creators to remix their uploaded songs. This new feature builds upon YouTube’s Dream Track, released last year, which enables users to create songs from text prompts and pre-recorded vocals. Artists like Charli XCX, Demi Lovato, John Legend, Sia, T-Pain, and Charlie Puth have already partnered with YouTube for this venture.
This AI tool acts as a co-producer, letting creators direct the remix process and generate 30-second samples based on their original song and text prompts. Imagine transforming a pop song into a reggaeton hit or creating a thrash metal rendition of the 1812 Overture – this tool makes it possible.
YouTube’s announcement explains that creators in the experimental group can choose an eligible song, describe their desired restyling, and generate a unique 30-second soundtrack for their Shorts. These AI-generated soundtracks will clearly attribute the original song through the Short itself and the Shorts audio pivot page. The AI contribution will also be clearly indicated.
This feature offers content creators a fresh way to soundtrack their Shorts, moving beyond pre-existing music libraries. It provides a unique audio experience specifically tailored to their video content.
Similar to other Dream Track features, the remix tool utilizes Google’s Lyria large language model. Lyria generates unique musical scores based on text prompts, analogous to AI image or video generation, but with audio output.
Google has confirmed that all AI-generated tracks will be clearly labeled, regardless of the level of human input. However, it’s unclear whether the company will apply its SynthID watermarking scheme to these new audio outputs.
