and
@neogunhero someone posted a video about this, so I put this answer there, as well...
I don't know if this applies to this tech, but everything that changes images with AI that I'm aware of has something called "denoising strength" that determines how close the AI has to adhere to the original. If you had an old image you wanted to clean up, but still wanted everyone in the image to be recognizable (or in art terms, keep the same art style), you would set this very low, like 0.1 to 0.3. To take a drawing, like I do, and make it photorealistic, you set that much higher. I usually go 0.6 to 0.8. After 0.8, AI just does whatever it wants, and you are just as like to get a picture of a zebra (this happened to me) as you are to get the original image.
My guess is that even if this doesn't have denoising strength (which would mean it isn't using diffusion) that there probably is an equivalent setting. They may have wanted to push it to photorealism believing people would be more impressed with this.