👐 Sora 2: OpenAI's Comeback to the Video Generation
OpenAI has released Sora 2, a new version of its AI text-to-video model.
The realism, sound quality, and camera work are truly impressive, even against strong competitors like Veo 3. The entire two-minute announcement was created by Sora 2 ⤴️
📸 Developers note that the model doesn't try to distort reality to follow the prompt. For example, if during the generation of a basketball shot the AI sees that the player misses, the ball doesn't teleport to the target to match the prompt, but rebounces off the backboard.
🎞 The startup also launches a standalone Sora app, a TikTok rival with an entirely AI-generated feed. With the cameo feature, users can place themselves in AI videos. You can make your image available to everyone, and also demand the deletion of any video with your AI double.
Currently, the app is only available for iOS users in the U.S. and Canada. Access to it, as well as to the model itself, is free but invite-only. Moreover, invites are required even for Pro subscribers who pay $200/month. They will also be able to use the Pro version of Sora 2.
🏆 The launch of the first Sora at the end of last year didn't go as well as OpenAI had hoped. After a delay of more than six months, the model struggled to stand out, and competitors like Google and Kling pushed Sora down the video generator rankings.
Now OpenAI is aiming for a comeback. The company has a strong product and a clear plan to promote it. Users are encouraged to keep their AI videos within the startup's own platform instead of sharing them elsewhere.
What do you think of Sora 2?
❤️ — amazing, it's a revolution!
🤔 — doesn't look like a breakthrough...
🔥 — I hate AI video at all
