OpenAI launches Sora 2 and an AI video social network
Published on
After the letdown that was the first Sora launch, OpenAI just released Sora 2, aiming to redeem itself (and to repeat the viral success of the Studio Ghibli-style images).
For those of you who are curious, below is a video they published at launch, fully generated with Sora 2 (both the video and the audio), and it really left me open-mouthed in many moments.
Let’s get to the details of what was announced:
the model itself took a leap in quality, maybe getting closer to (or even better, in some cases) the level of the Sora that was announced more than a year ago (which wasn’t the version of Sora that actually reached the public).
videos can now run from 5 to 10 seconds, with a clear improvement in object physics and in the ability to handle more complex scene changes.
the real turning point is the addition of audio. That includes both sound effects and dialogue that match the movement in the generated videos.
It’s a gigantic step pushing us even deeper into the “uncanny valley”, and it opens up a huge range of possibilities for storytelling.
Along with the model, the thing (which could be a flop or a stroke of genius) is that OpenAI also launched a new app, social network style, where you can use Sora 2 to generate videos and share them with your friends.
This app is focused on a feature called “Cameos”, which lets you record yourself (or upload a photo) and insert yourself into any AI-generated scene. The bet here seems to be on virality, and it has everything to become a new tool for “viral memes”, but the question is whether the app will have the legs to go beyond the initial novelty and not be just one more place full of “slop” content.
The app is launching for free (with usage limits) in the US and Canada. Pro subscribers will get access to a “Sora 2 Pro” model, and API access should arrive soon.
In my opinion, the model part (Sora 2 with audio) is what’s really impressive and has huge creative potential. The videos still look worse, overall, than the ones made by Google’s Veo 3 (Google DeepMind), but the quality has undeniably improved a lot. The social app part with “Cameos” feels more like an attempt to create a viral product, TikTok style maybe, and we’ll see if it sticks.