Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Why Sora Failed: $15M/day inference cost vs. $2.1M lifetime revenue

revolutioninai.com

Post reply on HN
Print view View original

41–50 of 65 posts

Re: Why Sora Failed: $15M/day inference cost vs. $2.1M lifetime revenue

#41

Post by northernsausage »

keiferski wrote: post #15

I would be curious to know if there is actually as much business economic demand for AI video compared to images (logos, product graphics, etc.) or text (blogs, content everywhere, etc.) My impression is that video is too complex to easily fit into an AI pipeline. Either you need something highly specific, like your own product’s UI. Or you need something personable and consistent, like someone talking into his camer…

As a product photographer/videographer - No its not good enough to understand products so each scene its different, you can't storyboard or collaborate with it,. For high end products (where the money is), colour shape, scale matter and its just not consistent enough for professionals. For cheap tiktop slop products is fine because what arrives it never what you ordered anyway.

The files are a pig to try and edit as well, making them beyond the generation and prompt costs expensive. At that point you might as well go and just film the ad.

Re: Why Sora Failed: $15M/day inference cost vs. $2.1M lifetime revenue

#45

Post by zarzavat »

brazzy wrote: post #30
PunchyHamster wrote: post #11

Earlier quoted context omitted.

The "matter of time" is getting more and more expensive, not cheaper, at least for next 2 years

Not sure what you're referring to. If you're talking about inference cost for frontier models, that's going up because researchers keep pushing those frontiers, often without considering cost. And while they're subsidized (to gain market share), users have no reason NOT to use the crazy expensive frontier models. Once the market consolidates, and users get used to the idea of using models that are "good enough" becau…

There's not much profit in inference, it's heavily commoditized. There is an illusion of potential profitability because the closed-weight models are currently a step ahead of the open-weight models. However, if you ignore the closed-weight models, then the open-weight models are also getting better every year. In the limit, the open-weight models will end up just as good as the closed-weight models.

AI is an inverse gold rush, the people who are getting rich off it are the people using it. The shovel-sellers are screwed.

Re: Why Sora Failed: $15M/day inference cost vs. $2.1M lifetime revenue

#46

Post by boredhedgehog »

keiferski wrote: post #15

I would be curious to know if there is actually as much business economic demand for AI video compared to images (logos, product graphics, etc.) or text (blogs, content everywhere, etc.) My impression is that video is too complex to easily fit into an AI pipeline. Either you need something highly specific, like your own product’s UI. Or you need something personable and consistent, like someone talking into his camer…

The aspiration is to replace the movie industry. That's a lot of demand.

Re: Why Sora Failed: $15M/day inference cost vs. $2.1M lifetime revenue

#48

Post by keiferski »

boredhedgehog wrote: post #46
keiferski wrote: post #15

I would be curious to know if there is actually as much business economic demand for AI video compared to images (logos, product graphics, etc.) or text (blogs, content everywhere, etc.) My impression is that video is too complex to easily fit into an AI pipeline. Either you need something highly specific, like your own product’s UI. Or you need something personable and consistent, like someone talking into his camer…

The aspiration is to replace the movie industry. That's a lot of demand.

But demand from whom? I feel like the biggest moneymakers in that industry are explicitly anti-AI.

General business stuff like content or images has demand from across the economy. “Replace Hollywood” is kind of a niche thing.

Re: Why Sora Failed: $15M/day inference cost vs. $2.1M lifetime revenue

#49

Post by danbruc »

How many seconds of video did they generate per day for those $15,000,000, i.e. what would it actually cost me to generate, say, a three minute music video for my garage band? This should probably take into account how many attempts I would likely need to arrive at something I am satisfied with.

Re: Why Sora Failed: $15M/day inference cost vs. $2.1M lifetime revenue

#50

Post by rsynnott »

santiagobasulto wrote: post #2

But doesn't $15M/day of inference cost imply "demand" from users? If this is the case, it's just a matter of time until costs can be reduced.

Costs would have to be reduced about 2,000 times just to break even, assuming that inference was the only cost, which of course it was not.
Post reply on HN