One of the most highly anticipated AI tools has officially launched: OpenAI’s AI video creator, **Sora**, made its entrance on Monday during the company’s **[12 Days of OpenAI](https://www.youtube.com/watch?v=2jKVx2vyZOY)** event.

OpenAI has previously hinted at Sora’s features, but how does the final version stack up against earlier glimpses? The organization has clearly put substantial effort into fine-tuning and augmenting its AI video creator in preparation for its public rollout.

Tech influencer Marques Brownlee provided a **[first glimpse of Sora](https://www.youtube.com/watch?v=OY2x0TyKzIQ)**, releasing his video review just hours prior to OpenAI’s official unveiling. So, what’s the conclusion?

—

### Strengths of Sora

As per Brownlee, Sora is proficient in generating lifelike environments. The AI-produced clips of natural scenes or famous sites—reminiscent of aerial, drone-like footage—often appear indistinguishable from genuine stock videos. However, as Brownlee observes, those who know the intricate details of a landmark may detect minor inaccuracies. Nevertheless, for the typical viewer, these videos don’t scream “AI-generated.”

One of Sora’s notable advantages is its capability to create **abstract visuals**. Whether it’s background graphics or screensaver-type animations, Sora executes these tasks with remarkable accuracy, even when given precise guidance.

Interestingly, Sora also shows decent performance with certain kinds of animated content, such as stop-motion or claymation-themed videos. The sporadic jerky motions typical of AI-generated videos can sometimes come off as deliberate, echoing the stylistic peculiarities of these animation forms.

What is perhaps most unexpected is Sora’s proficiency in producing **animated text graphics**. While numerous AI models encounter difficulties with text—frequently generating jumbled or absurd results—Sora can create clear, correctly spelled text for particular applications, like title screens. This marks a significant improvement over previous versions.

—

### Areas Where Sora Lacks

Despite its merits, Sora continues to face several of the same issues that have affected earlier AI video creators.

One primary concern is **object permanence**. Brownlee indicates that Sora has difficulties consistently showing objects during a video. For example, an item held by a person might inexplicably shift or disappear. This is indicative of the larger problem of AI “hallucinations,” where the model produces inconsistent or illogical outputs.

Another key limitation is Sora’s management of **physics and motion**. Creating photorealistic videos is still a challenge, as Sora often struggles to replicate natural movements. For instance, a person moving might suddenly accelerate or decelerate in an unnatural fashion. Body parts or objects can unpredictably distort or transform into completely different items, disrupting the illusion of realism.

While Sora has improved in generating specific text graphics, it still experiences challenges with background text, such as words on street signs or building surfaces. These often appear muddled or misspelled unless explicitly detailed.

—

### An Ongoing Journey

OpenAI recognizes that Sora is still in development. While it signifies progress compared to other AI video creators, it’s evident that certain obstacles—like sustaining object permanence, realistic physics, and accurate background text—remain significant challenges for all AI video frameworks.

At this stage, Sora presents promising opportunities for abstract visuals, landscapes, and specialized animations, but it does have its drawbacks. As OpenAI continues to refine the technology, observing how Sora develops in the upcoming months will be intriguing.