Today marks the launch of Sora, OpenAI’s highly anticipated AI video generation tool — though it’s initially rolling out to a select group of users. The excitement around Sora has been building, and early access reviews are already surfacing. Among them is a detailed first look by renowned YouTuber Marques Brownlee, who gave his initial impressions in a 15-minute review on his channel.
Sora represents OpenAI’s entry into the dynamic field of AI-powered video creation, offering a standalone platform rather than being integrated into OpenAI’s existing products like ChatGPT. For now, Sora lives independently at Sora.com, which features a curated scroll of recently generated videos. This homepage allows users to explore AI-generated clips created by others, bookmark their favorites, and analyze the text prompts behind each video. While the platform wasn’t live for general users at the time of Brownlee’s review, his insights provide an intriguing glimpse into what Sora has to offer.
What is Sora Capable Of?
At its core, Sora allows users to create videos from text prompts or uploaded images. Users can even edit existing Sora-generated clips, showcasing its versatility. The platform also introduces unique features that aim to push the boundaries of creative expression:
- Re-mix Functionality: This tool lets users tweak a video by describing specific changes they’d like to see. For instance, you could request a change in color scheme, a different animation style, or alterations to the characters or scenery. Sora’s “strength” setting enables users to control how drastic the changes should be, with higher values allowing more creative liberties.
- Storyboard Feature: For those looking to create cohesive, multi-scene content, the Storyboard feature allows users to link prompts together to produce sequences or entire stories. This is a big step forward in addressing one of the biggest challenges for AI video generators — maintaining consistency across multiple clips.
- Resolution Options: Sora supports resolutions up to 1080p, though generating higher-quality footage takes considerably longer. According to Brownlee, a 1080p video typically takes a couple of minutes to generate under current usage conditions. For faster results, users can opt for 480p or 720p resolution, which are generated significantly quicker.
Image Credits: Marques Brownlee
Strengths and Shortcomings
While Sora’s capabilities are impressive, they’re not without limitations. As Marques Brownlee highlights in his review, Sora faces challenges that are common to many generative AI tools:
- Object Permanence Issues: Videos created with Sora sometimes struggle with maintaining logical continuity. For example, objects might disappear and reappear without cause, or pass in front of and behind one another in ways that defy physics.
- Problems with Legs: Whether it’s people or animals, Sora has difficulty animating legs accurately in longer walking sequences. Brownlee pointed out instances where the front and back legs of a character swapped places in anatomically impossible ways, creating a disjointed and unnatural effect.
Image Credits: OpenAI
Despite these challenges, Sora shines when used for simpler tasks such as title slides, abstract animations, and stop-motion-style sequences. It’s less suited for creating photorealistic videos, with Brownlee noting that most clips “look AI-generated” on close inspection.
Built-In Safeguards and Ethical Considerations
OpenAI has implemented robust safeguards to ensure Sora is used responsibly. The platform prohibits content involving minors, violence, explicit themes, or anything that could infringe on copyright. Furthermore, Sora won’t generate videos based on images of public figures, recognizable characters, or branded logos.
Every Sora-generated video comes with a visual watermark to promote accountability, though Brownlee noted that the watermark could be easily cropped out. It remains to be seen how OpenAI will address this potential loophole in the future.
Performance and Usability
Sora’s speed and user-friendliness are commendable for an early-stage tool. During his testing, Brownlee found that generating a 1080p video took only a few minutes, though he cautioned that times could increase as more users join the platform. The clean and intuitive interface on Sora.com allows users to organize their projects into folders and save favorite clips for later inspiration.
Who Is Sora For?
According to Brownlee, Sora is an excellent tool for creators looking to experiment with short animations, abstract visuals, or artistic-style footage. It’s particularly well-suited for:
- Marketing Content: Think title slides, transitions, and unique animations for presentations or videos.
- Creative Projects: Abstract art, motion graphics, or stylized animations that don’t require photorealism.
- Experimentation: With its Re-mix and Storyboard features, Sora offers a playground for testing ideas and exploring new visual concepts.
However, it’s unlikely to replace professional video production tools anytime soon, particularly for tasks requiring photorealistic visuals or intricate scene coherence.
The Verdict
Sora marks an exciting step forward in AI-generated video technology. While it still has kinks to iron out — from object permanence issues to the quirks in animating movement — its creative potential is undeniable. With features like Storyboard and Re-mix, OpenAI is clearly positioning Sora as a tool for creative exploration rather than strict realism.
“It’s impressive that it’s AI-generated video,” Brownlee concludes. “But you can tell pretty quickly that it’s AI-generated video. Things just get really wonky.”
Sora is launching today for early-access users, with a broader rollout expected soon. If you’re a content creator eager to experiment with cutting-edge technology, Sora might just be the tool you’ve been waiting for. Whether you’re crafting abstract visuals or short animations, this AI video generator is worth keeping an eye on.
