Talk:Sora (text-to-video model)

related links

Hi, I have (I hope) added a proposed article for "Stable Video Diffusion" ; if and when accepted, it would be helpful to add that reference as a link to this article. 176.206.241.168 (talk) 23:18, 16 February 2024 (UTC)

Semi-protected edit request on 21 February 2024

Sora is not only a text-to-video model, it also does image-to-video and video-to-video. from openai.com/sora: "the model is able to take an existing still image and generate a video from it, animating the image’s contents with accuracy and attention to small detail." & "The model can also take an existing video and extend it or fill in missing frames." 209.53.181.216 (talk) 18:34, 21 February 2024 (UTC)

 Done - I added this to the lede, good catch Jamedeus (talk) 22:03, 21 February 2024 (UTC)

Inevitable legal issues discussion?

There should be some discussion--perhaps an entire section--about the inevitable legal issues with this model. How will companies be able to use this video model for commercial purposes without getting sued? Last I heard, OpenAI has not been obtaining permission from the myriad content sources used to create the Sora videos. ArthurRobertRobert (talk) 19:29, 24 October 2024 (UTC)

Do you mean here on the talk page or in article space? For the latter, absolutely not, unless there are sources that have had that discussion, and that aren't just the usual speculative dreck. I don't think there's any issue with a discussion here, as long as it's focused on discussions of same within reliable source, rather than editor speculation. Generally speaking, WP:CRYSTALBALL applies. cheers. anastrophe, an editor he is. 03:00, 25 October 2024 (UTC)

Fine, wait for the sources then. It certainly won't take long once this thing is launched. But we already have stuff like this going on https://news.bloomberglaw.com/litigation/openai-sued-over-using-youtube-videos-without-creators-consent

So... yeah. Call it "crystal ball" if you wish, but it's already going on, and there will be more. Turner.john60 (talk) 01:20, 29 October 2024 (UTC)

Then work up a section, embed the citations, and publish it. cheers. anastrophe, an editor he is. 01:39, 29 October 2024 (UTC)

This source doesn't even mention Sora, speculating that this data is for Sora would be WP:OR. I'm not sure how transcripts would be useful for a video model anyway unless they're also retaining the videos - sounds more like text-to-speech training data or just more text for LLMs. Jamedeus (talk) 02:09, 29 October 2024 (UTC)

Yeah, I wrote that in the vein of 'put up or shut up'. I have low tolerance for "I have an axe to grind, now you other editors, add this that and the other that I'm unwilling or unable to do myself'. cheers. anastrophe, an editor he is. 03:14, 29 October 2024 (UTC)

Yes, I agree, anyone who thinks there should be "perhaps an entire section" about something should write that section. Asparagusstar (talk) 03:53, 29 October 2024 (UTC)