The ‘Wild West of the Internet’ Returns

OpenAI recently launched their latest image and video generation model, Sora 2, a tool enabling the production of images and video from a text prompt. The initial Sora model debuted in February 2024, and had the capability to generate short clips without any sound. The output was limited to a few seconds in length, or up to 20 seconds for the highest paid subscription. While the earliest models did not produce very high quality clips, the model technology has advanced quickly. At this point, it is able to generate high quality videos with audio up to a length of about 10 seconds.

I attempted to get a Sora 2 invite code but I was unable to find one, as their service is not yet easily accessible. However, I was able to watch videos generated by other users and the video and sound quality is excellent. The newest models require powerful graphics processing units (GPU chips) to run the models on OpenAI’s servers, mainly H100 and H200 GPUs from NVIDIA.

Within the first week of its release, users created videos with copyrighted characters and logos, which quickly became a liability for OpenAI. Their initial policy was to allow any content to be generated unless specified as not allowed. This initially included anything against their guidelines or content which had been specifically requested to be blocked by copyright holders. After facing the threat of copyright lawsuits just days after release, they changed their policy to be inclusive, meaning corporations must specify that their images or intellectual property be allowed for generation. However, users are still finding ways to bypass these restrictions. This could be a resurgence of the ‘90s’ “Wild West of the Internet” phase, albeit in a new domain.

Image credit to 9to5mac https://9to5mac.com/2025/10/04/sora-2-clones-start-flooding-the-app-store-worldwide