Want to create your own music for YouTube Shorts? Now, this has become a possibility as Google’s DeepMind has collaborated with YouTube to create an AI-based music generation model with assistance from another set of tools that helps with the creative process. The model is built on Lyria: Dream Track and is in an experimental stage now.
All it took Google was eleven months after it announced the making of soundwaves through an AI-based music creation software that created tunes around word prompts. Now, using these tool sets, one can create music for YouTube Shorts and build out a tune from any snippet. Google will also be adapting SynthID, used to mark AI images, to its music creation tech.
The right beats, but the wrong timing?
However, the timing of Google’s efforts at making music using AI is steeped in irony, given that it could further heat up the debate around its use around creative arts. One where the actors and creative folk from Hollywood (Screen Actors Guild) struck work over commoditizing their work and their personas.
Several instances of AI-led deepfakes and voice usage resulted in the Screen Guild going on a long 118-day strike. As part of the settlement, the guild said it secured protections around the use of AI, including informed consent and compensation in cases where members are replicated digitally using AI.
While DeepMind has taken a step forward to ensure transparency over AI-generated music, the challenge would be for it to collaborate with YouTube to create tech that helps such music remain credible to the original creators. Not to mention the prerequisite of ensuring that the AI-generated replica remains aesthetically pleasing.
AI-generated music still needs to be real
In the past, music aficionados had complained that AI-generated music, especially Google’s experiments earlier this year, suggested that the longer one heard it, the more distorted it sounded. DeepMind has since clarified that this is partly due to the complexity of data being used in music models that cover notes, beats, harmonies and much more.
“When generating long sequences of sound, it’s difficult for AI models to maintain musical continuity across phrases, verses or extended passages. Since music often includes multiple voices and instruments at the same time, it’s much harder to create than speech,” DeepMind noted in a blog post.
DeepMind is releasing the applications in short pieces, initially rolling out a small set of creators to build 30-second AI-generated soundtracks in the voice and musical style of artists such as Charlie Puth, Demi Lovato, John Legend etc. So the users enter a topic, choose the artist’s track with lyrics. These are then backed with the voice of the musician to create new music. DeepMind shared this video as an example.
Artists are helping DeepMind fine tune the output
The two collaborators from the Alphabet stable clarified in no uncertain terms that all the artists listed are actually involved in the project and were helping test them out. Dream Track is getting a limited release now while Music AI tools will roll out over the next month or so.
On its part, the company released teasers that covers creating music in a specific instrument, creating one using the whole instrumentation set based on humming a tune, using chords that build on a simple MIDI keyboard to create an ensemble or choir and building backing and instrumental tracks for a vocal line that a user already has.
Just so that you’re aware, Google and Ghostwriter aren’t the only ones coming up with AI-led music tools. Meta has already sourced an AI music generator last June while Stability AI came out with one in September. Then there are others such as Riffusion that is raising money to up the ante around AI-generated music.
As for the creative industry’s peeve, well it looks like Big Tech and Big Money have gotten their way around it. Keep watching this space for more soundwaves.
