OpenAI CEO Sam Altman has just outlined two new directions for the Sora video generation model.
First is copyright control. In the future, rights holders will have more granular control over content generated with their characters, adopting an 'opt-in' model similar to portrait rights. Altman specifically mentioned Japanese creative content, noting that this 'interactive fan fiction' style of user participation holds unique value for rights holders.
Second is revenue sharing. Due to the unexpectedly high volume of user-generated videos, OpenAI plans to share revenue with rights holders willing to license their characters. How exactly? Altman says they're still experimenting but will iterate quickly and fix issues as they arise.
This is interesting. While most AI companies are still figuring out how to avoid lawsuits from rights holders, OpenAI is already thinking about how to make money together with them.
However, the revenue-sharing model sounds great in theory but is fraught with challenges in practice. How do you quantify the value of 'user-generated content'? Should it be based on usage frequency or the reach of generated videos? The balance of interests between rights holders and platforms has never been straightforward.
Altman's mention of 'quick iteration' roughly translates to: launch first, fix problems later. That's very much OpenAI's style.
As for copyright control, it's a more practical step. At least it gives rights holders a switch to prevent their characters from being misused. But how 'granular' can this control get? Can it restrict characters to specific scenarios, or even manage their expressions and movements? The details are still unclear.
For now, OpenAI is trying to transform Sora from a pure tool into an ecosystem connecting users, rights holders, and platforms. But this path has been tried before -- by Netflix, by YouTube -- with no clear success yet.
The only certainty is: the more AI-generated content flourishes, the murkier the copyright waters become.