The interesting thing about AI imagery is not just generating pictures from nothing. It is what you can now do with one decent real photograph.
Take a good product image and you can try a different background, place it in a room, extend the crop, create a different format for social media or explore colour and material variations. The original photograph gives the AI something real to anchor itself to.
That is particularly useful when the product is expensive to move, photograph or reshoot. You still need to check that the product itself has not quietly changed, because AI is very capable of “improving” details you wanted left alone.
But one good source image can now stretch much further than it used to.