WHAT WAS SORA TRAINED ON? CREATIVES DEMAND ANSWERS

Summary from the TLDR Design newsletter

OpenAI's Sora is an impressive diffusion model that translates visual data into "patches," or pieces, that the model can interpret and turn into hyperrealistic videos. While the company states it draws inspiration from large language models that acquire generalist capabilities by training on internet-scale data, it doesn't mention where the visual data came from. Many artists and creatives have already expressed on social media that the model was trained based on copyrighted data and will negatively impact their livelihoods.