OpenAI’s latest Sora release is a music video designed to capture the images a musician visualized in her mind while composing the piece.
Sora is an artificial intelligence video generator that is capable of producing multi-shot clips of a minute or longer from nothing more than a text prompt — but so far only a select few have used it to create content. OpenAI is working on security issues and slowly rolling it out this year.
One of the artists given early access to Sora is August Kamp, a musician, researcher and creative activist. She described Sora as representing a “turning point” for artists as it means the only limitation on visuals is the human imagination.
“Taking these pictures that I've held onto [in my mind] for two years and saying ‘August - we can share these with folks’. that's what I think is special about this tool,” she said.
What is the new Sora video like?
We don't have any specific details of the prompts used to generate the video or how many clips were required to create the full two minutes and 19 seconds.
Other video platforms generate up to three seconds for any single clip and up to 12 with extensions — but struggle with consistency beyond the first five seconds.
Sora changes that and is capable of creating much longer clips and often multiple consistent shots within each clip. In this case there seems to be a series of clips with 2-3 shots.
The new video, called Worldweight, is like a fever dream, taking you on a journey through a rain-soaked mind. The music is ethereal and gives off a sense of standing under a bus shelter on a dark evening in an empty street while the rain pours down in front of you.
Kamp said of the music: "I feel like this piece of art is my absolute heart and soul. I remember the way I felt when these notes tumbled through my hands and onto the keys of my synthesizer almost two years ago,” she said. Adding: “I remember how it rained outside and how I felt so lucky to be able to pour these feelings into a song — something that could hold onto them so I didn't have to anymore.”
When will Sora be available
OpenAI’s CTO Mira Murati hinted that Sora would be released this year. CEO Sam Altman has also said there would be multiple new models from the company this year, although didn’t specifically mention Sora.
Other models from companies like Runway, Pika Labs and Viggle are starting to catch up, adding features to make them more usable or improve the model. Even Midjourney is working on its own video AI model that could be released this year.
Being able to build and iterate on cinematic visuals this intuitively has opened up categorically new lanes of artistry to me.
August Kamp
With all of that I think it is very likely that there will be a gradual release of Sora to the public before the end of the year. It may start with opening it to high profile creatives, after all OpenAI seems to be actively courting Hollywood — but it all comes down to safety.
OpenAI has said it won’t release any model that could cause undue impact on the upcoming global elections at the end of this year, so if the company can’t tackle the misinformation risks caused by such a powerful video model before November — the release may be delayed.
Which is a shame as it sounds like Sora could bring new ways of making visual content. Kamp wrote: “Being able to build and iterate on cinematic visuals this intuitively has opened up categorically new lanes of artistry to me...I truly cannot wait to see what other forms of storytelling will come into reach with the future of these tools."
