Will there ever be a Sora 3?

[boorchess]

I wonder if open ai will release Sora 3, a sequel to sora 2. It should have all sora 2 features with new ones

[_j]

Sora was a system for OpenAI to set money on fire.

They could have asked me if I would ever upload a video of my face so that others can make brand-watermarked videos of me holding up a sign in a sharing-mandatory social media web site (or would use their browser), and saved themselves millions. (and I forgot: download an app for that).

The feature set:

@_j

Sora 2 as an app, with its “cameos” features (that had to be renamed), is about acting like a social media platform, and providing your own likeness and voice for others to deface within AI videos for their amusement, and also to ingest your identity into an identity verification system and then use it in AI model calls that can include training, and scrape your ChatGPT chat history for recommendations. Producing videos with bouncing watermarks.

Sora 2 on the API will reject any people seen in input reference images, people or faces or any character IP generated in output videos (which you pay for but don’t get delivered). You can get videos without watermarks that are actually suitable for something.

Sora (original, now a “turbo” version) can deliver videos with less watermark annoyance for “Pro” ChatGPT subscribers, different safety criteria, and was never offered on the API.

Try now as an API developer while it still remains - you are as likely to get your $6 for pro taken for a rejected input or output. Where a useful application needs input character reference and 100 calls to try to get your special effects shot without glaring symptoms, you cannot upload people.

OpenAI conspires to make sure partners have products rolled out with testimonials before you ever get API access, who are now hanging:

The original version of Sora also had pre-announcement insiders showing their long videos, but the product was only released in web form with short videos. It had aspirational timeline reference frames and video-appending features on the web that simply never worked well, besides people walking through walls and cars driving on rivers.

But most of all, the marketplace has rejected AI slop, simply being hated by consumers who also don’t like 24/7 roaring datacenters consuming water and jacking their power bills and PCs now unaffordable because chip production is bought out by AI companies. You even get in movie credits “no AI was used”.

So, giving a massively expensive product out to monthly subscribers is unlikely, as well as pro use, at least until AI video can be magically indistinguishable. OpenAI shut Sora down almost immediately for subscribers upon making a decision, with a very short API notice also, for a developed product. They have a killer app to use up all their computation in AI-powered coding, which people will actually pay for.

Any video product - once there is any niche that people are actually asking for and which can be fulfilled - likely will need to abandon the stink of even the name “Sora”. Magic 8-ball says all signs point to no.

[Riccardo_Melfi]

Yes, There will ever be a Sora 3.