OpenAI enters AI music generation race with new tool in development

Key Takeaways

OpenAI has embarked on developing a new artificial intelligence tool capable of generating music from text and audio prompts, marking the company's latest expansion into creative AI applications, according to a report published Friday by The Information.

The planned tool would allow users to create music for various purposes, including adding soundtracks to existing videos or generating instrumental accompaniment for vocal tracks.

The development represents OpenAI's most significant foray into music generation since the pre-ChatGPT era, when it released earlier models like MuseNet and Jukebox.

Training with classical music expertise

In an approach that distinguishes it from competitors, OpenAI has partnered with students from the prestigious Juilliard School to annotate musical scores, sources told The Information.

This collaboration suggests the company aims to incorporate structured musical knowledge into its training data rather than relying solely on audio recordings.

The use of annotated scores indicates OpenAI's focus on understanding musical elements such as notes, chords, rhythm, and form at a fundamental level.

This methodology could enable more precise control over generated compositions compared to systems trained exclusively on audio waveforms.

Details about the tool's release date and distribution format remain undisclosed.

OpenAI has not confirmed whether the music generation capability will launch as a standalone product or integrate into existing platforms such as ChatGPT or its video generation app Sora, which recently gained significant traction with over one million downloads in its first five days.

Entering a competitive and contentious market

OpenAI's music generation initiative places the company in direct competition with established players in the AI music space. Google has developed its Lyria model, while startups Suno and Udio have already gained substantial user bases with their music generation platforms.

However, the AI music generation field faces substantial legal challenges. In June 2024, major record labels, including Sony Music Entertainment, Universal Music Group, and Warner Records, filed copyright infringement lawsuits against both Suno and Udio.

The Recording Industry Association of America, representing the plaintiffs, accused the companies of training their AI models on copyrighted recordings without authorization or compensation.

In August 2024, both Suno and Udio acknowledged using copyrighted music in their training data but argued their practices constituted fair use under copyright law. The lawsuits remain active, with indie artists filing additional legal actions against the platforms in recent months.

The legal disputes underscore the complex copyright questions surrounding AI training data in creative industries. How OpenAI navigates these issues could significantly impact the viability and acceptance of its music generation tool.

Uncertain path forward

TechCrunch reached out to OpenAI for comment but had not received a response at the time of the original report.

The company's silence leaves key questions unanswered about licensing agreements, copyright protections, and safeguards against potential misuse of the technology.

While OpenAI has previously developed audio models focused on text-to-speech and speech-to-text capabilities, a full-scale music generation tool would represent a substantial shift in strategy.

The technology could democratize music creation for content creators, marketers, and amateur musicians while raising concerns among professional artists about AI-generated content flooding the market.

The global recorded music market, valued at approximately $28.6 billion in 2019 with streaming comprising 67 percent of revenue, represents a significant economic opportunity.

Read more: