OpenAI’s first Sora-generated music video released


The first music video created with OpenAI’s publicly unreleased text-to-video generator Sora has been released for viewing, and it’s very dreamlike.

The video, shared on YouTube, was made for chillwave musician Washed Out. The AI-generated clip is four minutes long and features quick zoom shots through different scenes, edited together to create a seemingly endless moving shot.

The video was created by writer and director Paul Trillo, who had early access to Sora in March. Trillo claims his creation is the “first official commissioned music video” developed with OpenAI’s Sora.

ADVERTISEMENT

The director said the music video consists of 55 clips stitched together for the premiere. After users on X inquired if he spent a lot of time editing the music video, Trillo explained he only made “minor touch-ups,” and he mostly did the cuts for the video.

“This was an idea I had almost 10 years ago and then abandoned. Finally was able to bring it to life,” Trillo said on X.

While Trillo claims to have created the first Sora-generated music video last month, the OpenAI’s model was used to make a video clip for an ambient track composed by indie musician August Kamp.

AI models capable of creating images, audio, and video have sparked discontent among creators, who feel their livelihood is threatened.

Last month, an open letter signed by more than 200 musicians, including Billie Eilish, Nicki Minaj, and Stevie Wonder, was published calling on AI developers, tech companies, platforms, and digital music services “to cease the use of artificial intelligence (Al) to infringe upon and devalue the rights of human artists.”

Meanwhile, OpenAI is facing several lawsuits for copyright infringement, including from comedian Sarah Silverman, writers George R. R. Martin, John Grisham, and Jodi Picoult, as well as the New York Times.

Sora is a large-scale AI model that transforms a user’s text prompt into 60 seconds of high-fidelity video. The service can generate videos from text prompts and pre-existing images or videos.

The model has not yet been publicly released, as OpenAI claims it’s being tested by red teamers to spot all of its possible weaknesses.

ADVERTISEMENT