>>61629
This is exactly why I was making the fox look like a cartoon.
When the fox looks photorealistic, the AI assumes that physics must apply, and will refuse to make things work, or cause absolute disasters.
Another big issue, is I have learned that the AI has not learned object permanence yet.
Which is why it has been so difficult to make it look as if something is being swallowed on screen.
Especially if you make reference to something being caused by something that should be seen
ex: "the fox goes down her throat" will create visuals of the fox visibly sliding down her throat...on the outside.
It also has to do with the AI being trained strangely on vocabulary.
Swallow and Gulp are two completely different concepts.
I have attempted to ask Chat GPT to enhance simple prompts into larger ones for a little more control,
and I have found out words like "shift" are interpreted as "transform"
others like "impossibly" or "unrealistic" seem to also make the subject transform into a monster...
I've literally been shotgunning prompts, using the fox to learn just how the AI interprets everything.
Message too long. Click
here
to view full text.