OpenAI Sora Launches Today: Key Highlights from the First Review

Sora, OpenAI’s new video generator, is officially launching today—though access is currently limited to select users. YouTuber Marques Brownlee broke the news in a video uploaded Monday morning, where he shared his initial thoughts after testing the tool during an exclusive early access period.

According to Brownlee, Sora operates on its own platform at Sora.com, which features a feed of recently-generated videos curated by OpenAI. (As of this writing, the site wasn’t live for TechCrunch.) Interestingly, Sora isn’t integrated into ChatGPT, OpenAI’s AI-powered chatbot platform, and appears to be a standalone product for now.

The Sora homepage allows users to bookmark videos for later in a “Saved” tab, organize them into folders, and view the text prompts used to create each video. According to Marques Brownlee, Sora can generate videos from both text prompts and uploaded images, as well as edit existing Sora-generated content.

With the “Re-mix” feature, users can describe the changes they want in a video, and Sora will generate a modified version based on their input. The “strength” setting in Re-mix determines how significantly the video is altered—higher values result in more creative liberties taken by the tool.

Sora supports video generation up to 1080p resolution, Brownlee noted, but higher resolutions take significantly longer to render. For example, generating 1080p footage takes eight times longer than 480p, the fastest option, while 720p requires four times more processing time.

According to Brownlee, generating a 1080p video typically took “a couple of minutes” during his testing. However, he noted that this was under light usage conditions. “That’s also, like, right now, when almost no one else is using it,” he said. “I wonder how much longer it’ll take when it’s open for everyone.”

Beyond single video clips, Sora features a “Storyboard” tool that allows users to link multiple prompts together to create scenes or sequences. This feature aims to improve consistency—a common challenge for AI video generators.

As for performance, Brownlee highlighted several shortcomings familiar to generative AI tools. Sora struggles with object permanence, resulting in objects that pass in front of or behind others in illogical ways, or that vanish and reappear inexplicably.

Leg movement is another weak spot. In videos featuring walking people or animals, Sora often confuses front and back legs, causing them to "swap" positions in ways that defy anatomical logic.

Sora includes several safeguards, Brownlee noted, preventing the creation of videos featuring individuals under 18, violence, “explicit themes,” or content that could infringe on copyrights. It also blocks video generation using images of public figures, recognizable characters, or logos. While each video is watermarked, the visual watermark can be easily cropped out.

So, what’s Sora best suited for? According to Brownlee, it excels at creating title slides with specific styles, animations, abstract visuals, and stop-motion-like clips. However, he cautioned against using it for photorealistic projects.

“It’s impressive that it’s AI-generated video, but you can tell pretty quickly that it’s AI-generated video,” Brownlee said, describing how most clips display noticeable quirks. “Things just get really wonky.”

Sora, OpenAI's text-to-video model, showcases innovative features like seamless video looping and the ability to extend clips forward or backward in time. These capabilities enable the creation of smooth, infinite loops and longer narratives without noticeable interruptions. Such features are particularly valuable for creators aiming to produce dynamic content or animations that feel cohesive and professional.

However, Sora has limitations that can hinder its usability in certain contexts. One significant challenge is its difficulty in accurately representing complex physical interactions, such as detailed motion dynamics or collisions. Additionally, while Sora can extend videos, scaling to longer durations (over a minute) often results in inconsistencies or degradation in video coherence. These issues highlight the model's struggle with maintaining both short-term and long-term temporal dependencies in generated content.

Furthermore, Sora's reliance on extensive training data brings risks of introducing biases from its training set. This can result in outputs that unintentionally reflect societal biases or other inaccuracies. These flaws, combined with the potential for misuse (e.g., creating convincing fake videos), make ethical considerations critical when employing Sora's technology.

For creators looking for precise, long-duration video solutions, these criticisms reveal areas where Sora may not yet be fully reliable, suggesting it is better suited for short, visually consistent outputs or experimental use cases rather than robust, professional-grade video production.

Cost and Accessibility

Sora by OpenAI is a groundbreaking text-to-video tool designed to democratize video production by significantly reducing costs. It allows creators to generate high-quality video content using simple text prompts. By eliminating the need for advanced equipment, large production teams, or extensive editing knowledge, Sora makes video creation accessible to smaller teams, independent creators, and even those new to video editing.

The tool can produce videos up to one minute long in high definition, offering features like animation from static images and filling in frames for seamless transitions. This streamlines workflows for marketers, educators, and content creators, reducing the need for expensive software or outsourcing tasks.

However, despite its strengths, Sora has notable limitations. The lack of physics consistency and object permanence can make videos appear unrealistic in certain scenarios, such as characters morphing or objects disappearing unnaturally. Additionally, its current accessibility is restricted in some regions like the EU and UK, likely due to regulatory concerns. These limitations may deter some users from fully adopting Sora for professional-grade content creation, as polished results still require manual tweaking​.