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.
> 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.
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.
A source familiar with the matter tells The Hollywood Reporter that Disney is also exiting the deal it signed with OpenAI last year, in which it pledged to invest $1 billion in the company and agreed to license some of its characters for use in Sora.
“As the nascent AI field advances rapidly, we respect OpenAI’s decision to exit the video generation business and to shift its priorities elsewhere,” a Disney spokesperson said. “We appreciate the constructive collaboration between our teams and what we learned from it, and we will continue to engage with AI platforms to find new ways to meet fans where they are while responsibly embracing new technologies that respect IP and the rights of creators.”
Also "exit the video generation business" seems somewhat notable, suggesting they're not just planning to launch a different video gen product to replace Sora?
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.
This data is pretty questionable. OpenAI employees have said on Twitter that it does not account for ChatGPT Enterprise, where most of their growth is, which is quote-only and not paid by credit card.
You have more info about the inflated token use? I’m using codex cli a bunch now, but the reported token usage seems like an order of magnitude higher than, say Claude code with opus.
Idk if it’s because I set codex to xhigh reasoning, but even then it still seems way higher than Claude. The input/output ratio feels large too, eg I have codex session which says ~500M in / ~2M out.
I wish I had hard evidence but it is mostly an observation. I do use Codex a lot and I felt a drastic change from like one-two months ago to this day.
It used to give me precise answers, "surgical" is how I described it to my friends. Now it generates a lot of slop and plenty of "follow ups". It doesn't give me wrong answers, which is ok, but I've found that things that used to take 3-4 prompts now take 8-10. Obviously my prompting skills haven't changed much and, if anything, they've become better.
This is something that other colleagues have observed as well. Even the same GPT5.4 model feels different and more chatty recently. Btw, I think their version numbers mean nothing, no one can be certain about the model that is actually running on the backend and it is pretty evident that they're continuously "improving" it.
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.
Seedance has a lot more restrictions now, but still arguably not as much, it's probably cheaper for ByteDance to run, and as you said, it at least looks good enough to be worth paying for.
Sora had all of the downsides, and attracted all of the scrutiny. Local-first is definitely the way.
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..
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.
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.
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.
Agree, and didn't intend to imply that. This is just a good startup move that gets a big headline because it's OpenAI. Other startups around the world do the same thing all the time.
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.
> 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
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.
> 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.
You know they are burning money dangerously when they decide to focus on the area in which they are getting their asses kicked...
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.
Didn't they cut a huge deal with Disney just 3 months ago?
https://openai.com/index/disney-sora-agreement/
Yes, and Disney is apparently no longer investing in OpenAI, making one more example of an OpenAI investment hype cycle that turned out to be hot air.
Disney Exits OpenAI Deal After AI Giant Shutters Sora
https://www.hollywoodreporter.com/business/digital/openai-sh...
Also "exit the video generation business" seems somewhat notable, suggesting they're not just planning to launch a different video gen product to replace Sora?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.
They are just cancelling side projects because Anthropic is dominating in enterprise and side projects (probably) don't make profit. https://x.com/ShanuMathew93/status/2031074311629353299
This data is pretty questionable. OpenAI employees have said on Twitter that it does not account for ChatGPT Enterprise, where most of their growth is, which is quote-only and not paid by credit card.
You have more info about the inflated token use? I’m using codex cli a bunch now, but the reported token usage seems like an order of magnitude higher than, say Claude code with opus.
Idk if it’s because I set codex to xhigh reasoning, but even then it still seems way higher than Claude. The input/output ratio feels large too, eg I have codex session which says ~500M in / ~2M out.
I wish I had hard evidence but it is mostly an observation. I do use Codex a lot and I felt a drastic change from like one-two months ago to this day.
It used to give me precise answers, "surgical" is how I described it to my friends. Now it generates a lot of slop and plenty of "follow ups". It doesn't give me wrong answers, which is ok, but I've found that things that used to take 3-4 prompts now take 8-10. Obviously my prompting skills haven't changed much and, if anything, they've become better.
This is something that other colleagues have observed as well. Even the same GPT5.4 model feels different and more chatty recently. Btw, I think their version numbers mean nothing, no one can be certain about the model that is actually running on the backend and it is pretty evident that they're continuously "improving" it.
I guess this is a bullish sign OpenAI has hired a lot of PMs from Google!
"OpenAI’s top executives are finalizing plans for a major strategy shift to refocus the company around coding and business users" - WSJ
Coding is where the money is. https://news.ycombinator.com/item?id=46432791#46434072
*Forcing AI devtools down everyones throat is where the money is.
Shovel selling and instruments to dismantle whats left of working class power.
Unlike, say, Seedance 2.0 (which has yet to come to the West), Sora 2 was more of a tech demo than anything usable:
* It was (assumedly) expensive to run.
* It was not good enough for customers to seriously pay for.
* There were too many content restrictions for it to be fun for most people.
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.
Seedance has a lot more restrictions now, but still arguably not as much, it's probably cheaper for ByteDance to run, and as you said, it at least looks good enough to be worth paying for.
Sora had all of the downsides, and attracted all of the scrutiny. Local-first is definitely the way.
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..
Sora was fun
But it was largely fun to try to transgress against the limitations. Who could trick the AI to generate something outlandish and ridiculous.
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.
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.
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.
> it ain't a masterful Xanatos gambit.
Agree, and didn't intend to imply that. This is just a good startup move that gets a big headline because it's OpenAI. Other startups around the world do the same thing all the time.
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.
Yes, sadly, the West is not the leader. The work done by the Chinese labs is just so damn good.
How much money did they burn on this? And for what? Nothing?
an official post
> 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
(https://x.com/soraofficialapp/status/2036546752535470382)