...to this.

From this...

...to this...

...and then with another step and some Photoshop work.

From this...

...to this.

With some work I could have gotten some appropriate bleeding, but I keep running into the problem where, if I ask for any blood at all, there's blood all over the place, and little or none coming from the right places.

From this...

...to this.

It's odd in these conversions how approptiate facial expressions are often lost. I find that if my prompt tries to fix things like that, it's at the expense of other things going wrong.

It seems to work much better to take an image like the one above and do I2I (image-to-image) processing again, this time with a prompt like:

Girl on the left looks down at her breast in horror. The girl on the right smiles with delight.

What are you using to get this images?

>>15837

I'm using Qwen Image Edit Rapid v1.0, running locally, of course, because content filters on most online services would block this content.

You can do a lot of this with no more than the simple prompt "Make real".

Oddly enough it's better at turning simple line drawings and low-detail artwork into realistic scenes. The more that realistic shading an detail that are already present, the more the model tends to give up and leave that stuff alone.

From this...

...to this.

I've tried modifying other images depicting amputation besides this one. It seems to be a common thing that the raw stump gets covered with a rimmed metallic plate. An odd kind of consistency there.

So, onto a little Photoshop...

...and with admitedly not a huge amount of work expended to do better, I at least got to this.

Here's a bit of classic Dolcett...

...and what Qwen Image Edit Rapid v1.0 did to make it look real.

Okay, I had to photoshop away a duplicate head that was still on the body of the woman on the ground after Qwen Image Edit Rapid v1.0 got done with the image.

I'm sure this must be sacrilege of some sort, but here's the mascot for the guro.cx Art forum...

...rendered semi-real by AI.