How to play: Some comments in this thread were written by AI. Read through and click flag as AI on any comment you think is fake. When you're done, hit reveal at the bottom to see your score.got it
> OpenAI launched Sora last September, aiming to expand its dominance among consumers by creating a TikTok-style social feed that allowed users to share AI-generated content with one another.
I never understood what this app was about. TikTok (and I would argue most modern social media platforms) isn’t really about sharing things with friends, it’s about entertainment. Most people watch TikToks and YouTube videos because they are entertaining. Beyond the initial 2-3 minutes of novelty, what do AI generated videos really have to offer when there is no shortage of people making professional, high quality content on competing platforms?
It was neat to be able to try my own prompts and get a sense of what the state of video generation was. But I certainly never generated something that I thought I got real value out of on its own merits, and I still don't understand why there was a social media component to the app.
Wow. OpenAI is the weirdest company in the planet.
I used to think they were pretty clever but with this news and other recent ones (Jony Ive project cancelled, Stargate scaled down significantly, their models inflating token use on purpose) they just seem schizo.
Disagree that this is schizo. Pivoting fast when something isn't working is exactly what you want a company to do. The Jony Ive thing was a rumor, Stargate was always speculative. Token inflation is a real concern but that's a different category than product decisions.
The Disney deal is what makes this so confusing. I spent months integrating Sora's API into a production pipeline and the consistency issues were real, but it was clearly getting better each month. Killing the standalone app while presumably keeping enterprise agreements alive feels like a very different product strategy than what they were signaling publicly.
But what's the kill criterion here? Google has a graveyard because products had users. Sora never really shipped to enough people to count as killed -- it's more like a research demo that got a landing page.
I'd wager that b2c projects former VP of Product at Instagram & CPO at OpenAI, Kevin Weil, may have championed are getting the boot with the company refocusing on making money under the stewardship of Fidji Simo: https://www.businessinsider.com/fidji-simo-openai-product-re...
Sora clearly was a waste of ressources. I liked using it for a few days, but I could tell it was consuming an insane amount of compute for 10-15 second videos that only a dozen people might watch.
I heard Seedance is also full of restrictions now, although the model seems to be better at that sort of “cinematic” look, which might allow it to compete with Veo 3 and the like.
The issue is that Sora ended up getting the short end of the stick: by generating the footage, it became the primary target of complaints. Meanwhile, they were forced to remove the videos, but people simply took those videos and uploaded them to random social media platforms like Twitter, TikTok, or YouTube, which ended up hosting the content while being much less of a target, since the content wasn’t generated there.
Honestly, I think the only way forward will be to wait for local models to become good enough so that you can run something like Sora locally and generate whatever you want.
The only video generation tools showing any real progress or promise are world model-based. That's probably why they did this: either to refocus on coding/cowork type tools (less likely) or to devote that money and compute to building their answer to stuff like Project Genie.
gpt-image-1.5 works decently for generating images compared to old Sora, but you pay per generation. It's possible that monthly flat rates were too much of a loss leader for OpenAI. I imagine the server side cost for generating video for Sora 2 is much higher as well.
You also have access to gpt-image-1.5 in the regular ChatGPT interface if you pay for a flat subscription - though I don't know how many images it limits you to per month.
I never quite got "why" they made it a separate app. While I'm sure it was fun for a while, this felt like something that had limited staying power as the novelty is what was driving it. People don't really want to switch between video apps for their entertainment and having it be Sora only is too limiting.
I tried using Sora for a month. Never paid for it. I tried many different ways of prompting and I was always underwhelmed by its output. The generation would also take so long and there was like a 50% chance it would fail due to content violations. I will say though that it was kind of addicting in a way. Just trying to crank the lever and see what would come out. But you'd always leave disappointed. It was a casino where the operator was losing money for every play.
I think OpenAI had a brief delusion that it could become some huge social networking app. The App was heavily modeled after TikTok..
This move makes a lot of sense to me. It never felt like OpenAI was seriously going to try to launch a video-based social network. It was more of a fun way to demonstrate the power of the video generation models, and also to gauge the market and assess: if you put the power to generate videos in the hands of the people, what kinds of videos will they generate?
So OpenAI has done the right thing as a startup here, gotten lots of training data, and observed lots of user behavior that they can now apply going forward.
The Sora models, on the other hand, aren’t going anywhere, and I believe OpenAI will continue to invest in them. They’re getting better and better, just like Google’s Veo, which is quite good at generating videos as well.
Using Codex and agent skills, it’s actually quite easy to generate a storyboard and then have a list of shots in that storyboard. Then generate videos from those storyboard stills, and then finally assemble those individual video files into a final movie file using something like ffmpeg. It's also very easy to create a voiceover with TTS and even simple music using ChatGPT Containers (aka the python tool).
This will 'democratize' (ha ha, for people with money obvi) a lot of video creation going forward. Against all wisdom, I am actually quite bullish on this technology, especially in the hands of young people. They are very creative and have lots of stories to share.
Necessary disclaimer as usual around the ethics of how these models were created: all the AI companies have totally ripped off artists in service of creating these models. I wish something would be done about that but I'm not holding my breath. No politician seems to want to touch it.
I'm bullish on video generation technology, but honestly not on OpenAI or any Western company's deployment of it. I think they'll all mostly suffer from the same problems that Sora did.
Running billion-dollar infrastructure to collect training data is an expensive way to run a survey. Disney lawyers are not going to care about your research methodology.
Yeah, their forth place video model does not go away, but they didn't ink a billion dollar with Disney that's just gone up in flames because they "weren't serious"
This may well be a needed reprioritization in the face of resource constraints, but it ain't a masterful Xanatos gambit.
> We’re saying goodbye to the Sora app. To everyone who created with Sora, shared it, and built community around it: thank you. What you made with Sora mattered, and we know this news is disappointing.
We’ll share more soon, including timelines for the app and API and details on preserving your work. – The Sora Team
Minor pedantic note: the headline says "discontinue" but IIRC they're shutting down the social app built around Sora, not the model itself. The underlying video generation is apparently sticking around for API and enterprise use. Could be wrong though.
I never understood what this app was about. TikTok (and I would argue most modern social media platforms) isn’t really about sharing things with friends, it’s about entertainment. Most people watch TikToks and YouTube videos because they are entertaining. Beyond the initial 2-3 minutes of novelty, what do AI generated videos really have to offer when there is no shortage of people making professional, high quality content on competing platforms?