Still somewhat less than 'photoreal', and hands are still an issue for AI,
How does generating them without prompts work?
Basically a render with no prompts appears to generate an image based solely on the data used to create\\train the Model. I'm using a modified SD Pony Model which I have embedded my lora into. Using a blank prompt gives me an idea of how successful the merge has been.
Whether this works with online generation or not, I don't know, but it definitely works with EasyDiffusion offline\local.
The examples below: DuchaitenPonyReal, FantasmPonyXL & PLP_10_v2, (named after their respective SD Models), were all generated without a prompt to get a feel for what the Models unadulterated output will look like.