Photo Credit: Michal Czyz
According to Bloomberg, YouTube originally planned to release the voice-generation tool at its Made On YouTube event. However, finalizing deals with record companies for the rights to popular artists’ voices is taking longer than expected.
YouTube is reportedly still finalizing licensing deals with the Big Three — UMG, SME, and WMG. Once the agreements are locked in, YouTube will use licensed artists’ voices to train their AI models, and the resulting AI-powered tool will allow creators to generate and use artists’ voices in their content.
According to Bloomberg, a rollout of the beta version of the tool would require individual agreements to cover the rights to voices. However, confidential sources have revealed that major record labels continue to negotiate the agreement terms before signing off on any deal.
In beta, YouTube’s new creator AI tool would be backed by licensing deals with major labels. Artists can choose to opt-in for YouTube’s training model, after which a select pool of creators would gain the ability to generate these artists’ voices for their content. After the tool is broadly released to all YouTube users and creators, a licensing agreement on the backend will protect them from claims of infringement.
Alphabet-owned YouTube has created several AI products over the years, but will require cautious navigation of legalities if it plans to follow through with this latest tool. Last month, YouTube released a new suite of AI tools for creators, allowing generation of backgrounds for content, and automated dubbing in multiple languages. The tools allow creators to expand their reach by refining content presentation, and monetize videos by targeting non-English-speaking audiences across multiple geographical regions.
For creators, monetization of content continues to be the primary objective. Alongside YouTube’s latest AI tools, companies like Identifyy (owned by HAAWK) are allowing creators to harness the advanced features of Content ID, so they can match and monetize their content across platforms like YouTube — and even Meta.
As AI gains popularity, all major labels have communicated their interest in embracing AI. But it’s a tricky landscape, and the music industry sees AI’s assistive aspect alongside the threat it could pose to copyright.
Naturally, music companies have been wary of the technology. To ensure they emerge on the right side of AI, labels have sought avenues that allow them to control the anatomy and power of AI — by creating frameworks that restrict developers from training their models on copyrighted works without adequately compensating rights holders.
In the past, piracy and user-generated content wreaked havoc on the industry before streaming services like Apple Music and Spotify stepped into the picture. The industry is still reeling from the notorious Ghostwriter AI-generated song releases from Drake and Weeknd. Such preposterous releases revealed generative AI’s prowess to the industry.
Now, while major music industry players set out rules and guidelines for AI, the industry is seeing rampant copyright infringement lawsuits. AI’s intersection with creative works that utilize likenesses to artists, their images, and their creations has become the bane of the industry’s existence. Streaming services have also readily accepted removal requests by labels for content that features AI-generated vocals that sound like popular artists.
With YouTube’s latest suite of AI tools and this reported voice generation tool that’s still in the works, YouTube appears to be positioning itself as a partner that allows labels to properly harness AI.
AI’s role within music industry processes now appears inevitable. Artists realize these models could be the key to novel avenues for creative expression and revenue.
However, sources reveal artists fear that participation in the YouTube tool would place immense power in the hands of creators — who could use their voices to create and sing lyrics, or make statements that are detrimental to their brand.
Other vital details to iron out include how YouTube would train the model and how artists would exercise the power to opt-in or stay out of the AI model training dataset. There’s also a question of whether artist compensation would arise from allowing the use of their voice as an input into the model, or from the output generated from the YouTube tool. Or, will they be compensated for both? Sources have hinted that YouTube is leaning toward paying a lump sum licensing fee to ensure simplification of payouts.
Labels are also keen to be seen as adapters and promoters of the latest tech and AI models — and have exhibited interest in being part of the innovations and progression in the AI realm.
But this motivation could possibly arise from fears of being left behind, or rendered obsolete in the face of rising AI tools and related developments. According to Bloomberg, YouTube is a reliable early partner to take charge of this space. And even though the deal terms are complicated, rights holders are keen to sign off on a deal. There are also talks of fears among labels that they ‘could get left behind’ if they don’t come to the table with licensing deals now.
In August, UMG and YouTube had announced they will jointly develop AI tools, announcing ‘AI music principles‘ to launch a ‘Music AI Incubator.’ With artificial intelligence content and unauthorized soundalikes making waves, the strategic deal culminated in an AI framework to control how AI affects artists, and ensures that artist compensation remains protected as the technology develops.
In a blog post titled An Artist-Centric Approach To AI Innovation on YouTube’s official Blog, UMG CEO Sir Lucian Grainge had emphasized how ‘music has always been fueled by dynamic exchange,’ adding, “There will surely be give-and-take as we work through AI’s opportunities and challenges. It’s this spirit of exchange — collaborative and competitive — that has always driven creative progress, and fills me with anticipation and optimism for music yet to come.”
