OpenAI confirms Sora will launch to the public this year

When OpenAI first stunned the internet with Sora in February, everyone had the same question: when do the rest of us actually get to touch it?

Now we finally have an answer. In an interview with The Wall Street Journal, OpenAI Chief Technology Officer Mira Murati confirmed that Sora will be publicly available “this year,” adding that it “could be a few months.”

Up until now, access has been strictly locked down to visual artists, designers, and filmmakers. But even with that tiny group, clips have been leaking all over X, leaving everyone else waiting in line.

Sound, editing tools, and an eye-watering power bill

Murati revealed that the current version of Sora is far from the finished product. OpenAI is planning to “eventually” integrate audio, giving those photorealistic scenes matching soundscapes.

More importantly, they are working on video editing capabilities. Murati noted that generative models don’t always get things right on the first prompt, so the team wants users to be able to tweak and remix Sora’s outputs directly.

Behind the scenes, though, powering Sora is a massive headache. Murati admitted the model is “much more expensive” to run than text or image tools. The company is actively trying to bring running costs down to roughly what DALL-E costs before opening the floodgates to the public.

New demos shared with the Journal showed off a mix of styles, including an animated bull wrecking a china shop and a mermaid reviewing a smartphone.

Shaky data questions and election-year guardrails

Things got noticeably awkward during the interview when Murati was pressed on what training data feeds Sora. She dodged specific questions, stating only that it relied on “publicly available or licensed data.”

When asked whether Sora trained on videos from YouTube, Instagram, or Facebook, Murati claimed she simply wasn’t sure. The only concrete source she confirmed was Shutterstock, with whom OpenAI already has a licensing partnership.

Safety is the other big reason the rollout is taking months. With the 2024 presidential election looming, the threat of realistic AI video misinformation is huge.

Murati said Sora will likely follow DALL-E’s safety policies by blocking users from generating public figures. OpenAI will also slap watermarks on outputs, though tech critics already know watermarks can be easily cropped or bypassed.

Why it matters

Bringing Sora to the general public turns text-to-video from a viral tech demo into an actual commercial battleground. If everyday creators can spin up sixty-second cinematic clips from a laptop, the barrier to entry for video production drops off a cliff.

Still, we think users should manage their expectations for the initial rollout. Until OpenAI nails granular editing controls and adds native audio, Sora remains more of a storyboard inspiration tool than a true replacement for professional video pipelines.

OpenAI’s real vulnerabilities look to be financial and legal. Running full-fidelity video generation for millions of free or cheap tier users could bleed immense amounts of cash unless their optimization works miracles.

At the same time, Murati dodging questions about YouTube training data suggests copyright battles are far from settled. Restricting public figures helps sidestep election scandals, but keeping Sora cheap, legally safe, and genuinely useful all at once will be a tightrope walk.