When OpenAI first launched Sora in February 2024, it was a rough, but revolutionary opportunity for users to create their own videos from a single prompt, much like Veo 3, Runway and other best video generators.
Today’s launch of Sora 2 is a leap forward; OpenAI promises the updated video generator is now even more aware of how objects and people move in the real world and says it can generate videos far better than before.
But those eager to test the new model may have to wait indefinitely. While the launch of Sora 2 includes a new iOS app designed to make AI video collaborative, social and fun, users must be invited to Sora 2.
How to get on the waitlist
OpenAI is rolling out access gradually:
- Invite-only first: Some early users will receive direct invites, each with a limited number of friend passes.
- ChatGPT users next: In the coming days and weeks, OpenAI will prioritize Sora 1 power users, ChatGPT Pro subscribers, then Plus and Team plans. If you’re eligible, you’ll get a push notification in the ChatGPT app.
- App + web: The Sora iOS app launches tomorrow in the U.S. and Canada, with more regions to follow. You can also try sora.com directly.
- Developers: API access and Sora 2 Pro are coming soon.
If you want in, the best move is to ensure you’re subscribed to ChatGPT Pro or Plus and watch for the invite notification. Otherwise, you’ll need to keep an eye out for a friend pass.
What’s new in Sora 2
The second generation promises improvements that make AI video more convincing and versatile:
- Better physics and consistency: Characters and objects move more naturally across multiple shots.
- Audio + video sync: Voices and sound effects line up with visuals instead of drifting apart.
- Rich world modeling: More faithful environments for experimenting with AI “world simulation.”
The Sora app: video creation goes social
OpenAI has a new initiative with Sora 2, one that encourages collaborative generation. With 10-second video generation (that’s 2 seconds more than Veo 3), here’s what’s possible within the app:
- Create videos from text prompts or inspiration images.
- Cameo yourself or friends (with their consent) using secure likeness + voice capture.
- Remix videos in “chains,” like TikTok duets.
- Follow and discover new creators.
- Keep creations private until you decide to publish.
This “AI-first social video app” framing is new territory for OpenAI, and it’s what the company hopes will make Sora 2 a success, not just another technical milestone.
Guardrails around identity and safety
The Cameos are fun, but they raise a slew of questions about how OpenAI intends to keep the identities of its users safe. Here’s how the company intends to do that:
- Cameos are opt-in only and verified with video + audio capture.
- Public figures can’t be generated unless they’ve uploaded their own cameo.
- Every video carries C2PA provenance metadata and visible watermarks.
- Parents can configure teen accounts with content filters and DM restrictions.
OpenAI is hoping that the combination of provenance, watermarks and parental controls will be enough to get ahead of the inevitable deepfakes, but only time will tell as the model rolls out and users find workarounds.
The takeaway
With the success Google has seen from Veo 3, which is arguably the best video generator available, Sora 2 is entering a competitive market with what the company hopes to be an innovative solution to stand out, merging creation, identity and sharing into a single platform. That could open the door for everyday users, not just video producers or expert content creators, to cameo in skits, remix friends’ clips or scroll a feed of AI-made shorts.
If GPT-4 helped ChatGPT go mainstream, Sora 2 could be the moment AI video takes off the same way.
More from Tom’s Guide
Back to Laptops
