Leo's Lightbulbs💡

The big news this week is OpenAI’s Sora video generator being leaked. Honestly, I’m surprised it didn’t happen sooner. For those just catching up, OpenAI created their own video generator model to compete with Runway, and seemed to really go multiple steps farther than anyone else. However, they kept it in closed beta for months and months (since February 2024) and basically every other video generator out there caught up and even surpassed them. In particular, video is one of the only places where Chinese companies have been able to keep up with American ones, for whatever reason (Kling/Minimax being the champ of the bunch).

Apparently, a group of artists who signed up to use Sora, a video generator, later decided to pull a PR stunt by making the model open to anyone. Why did they sign up to test Sora if they held such strong views? Did their views change over time?

Regardless of their reasoning, I presume this will push OpenAI to release Sora sooner. Personally, I think that image and video AI democratize access and opportunity for creators, and I’m looking forward to seeing what comes next!

Bluesky also made the news for it’s incredibly open API making it a prime source for AI training data. If you don’t want your data scraped, for now the only remedy is to avoid Bluesky.