The numbers in the brackets are the random number seeds assigned to each image in a stack
The way the LD model works is it starts with a noise pattern that uses that value. Then it resolves that noise to the image you selected. The really cool part is. When you use the same prompt and seed you end up with the same image.
For example :
!dream "film still of John Cena as Nada in Bank scene in They Live 1988" -W 704 -S 1225724671
Will produce this image. Try it.
Now, it's close, but what's he doing? He's not in the scene. He needs a shotgun and sunglasses!
!dream "film still of John Cena as Nada wearing sunglasses and holding a shotgun with a bandolier in Bank scene in They Live 1988" -W 704 -i -S 1225724671
So by editing the prompt to add sunglasses and a shotgun we get:
But in the movie he's wearing a flannel shirt, not a t-shirt
So we can add that to the prompt
!dream "film still of John Cena as Nada wearing a flannel shirt and sunglasses and holding a shotgun with a bandolier in Bank scene in They Live 1988" -W 704 -i -S 1225724671
And we get :
We successful put John Cena into They Live! But wait, What if I want Dwayne Johnson...he's also a modern wrestler.
Just edit the prompt again:
!dream "film still of Dwayne Johnson as Nada wearing a flannel shirt and sunglasses and holding a shotgun with a bandolier in Bank scene in They Live 1988" -W 704 -i -S 1225724671
And maybe Dave Bautista:
or Randy Macho Man Savage:
who said it has to be real people, how about Kratos?
You can do a LOT with seed editing.