Toys’R’Us released a TV ad generated with Sora a few months ago. Everyone loathed it and the ad caused massive brand damage. Toys’R’Us executives considered it a huge success. [YouTube; NBC] You ne…
I don’t know the actual budget, but I think it probably cost much more than 1kUSD, though probably still less than real human work would cost.
It’s important to note that no shit could last more than a second or two because after that the generated video starts to much more noticably have errors. So at minimum you still need editors (plus the music needed to be composited, etc). Also, as the article notes, all the logos needed to be added in post as well because GenAI cannot reliably do text or logos. With that in mind, I’d guess there was probably a significant amount of “cleaning up” that had to be done in post as well.
With all that said and done, I’m sure the commercial was not exactly dirt cheap, but it WAS probably still cheaper than having dignity and paying humans.
What’s actually kind of wild, though, is a lot of these shots just look like bland stock imagery. And since they couldn’t have any cohesion between shots because of GenAI’s own limitations, the majority of these shots could have been replaced with stock footage and they probably would have only needed to CGI a few different shots…
I don’t know the actual budget, but I think it probably cost much more than 1kUSD, though probably still less than real human work would cost.
It’s important to note that no shit could last more than a second or two because after that the generated video starts to much more noticably have errors. So at minimum you still need editors (plus the music needed to be composited, etc). Also, as the article notes, all the logos needed to be added in post as well because GenAI cannot reliably do text or logos. With that in mind, I’d guess there was probably a significant amount of “cleaning up” that had to be done in post as well.
With all that said and done, I’m sure the commercial was not exactly dirt cheap, but it WAS probably still cheaper than having dignity and paying humans.
What’s actually kind of wild, though, is a lot of these shots just look like bland stock imagery. And since they couldn’t have any cohesion between shots because of GenAI’s own limitations, the majority of these shots could have been replaced with stock footage and they probably would have only needed to CGI a few different shots…
What if I told you that the goal was not to make a good product, but to increase shareholder valuation?
In that case, all the “problems” disappear in light of chasing after the singular goal.