After hearing two eyewitness accounts of the same accident, you begin to wonder about history.
Unknown
Using AI to generate images shows just how many variants are possible from a single image description. For example, Gab AI (https://gab.ai or https://gab.ai/start/gabby) did generate the following image:
If you would describe the image over the telephone without video, how would you describe it? I mean, sure, it’s a young woman with light red hair, a green camisole leotard with a short plain shirt. But even if you provide this information, there are countless variants possible.
I gave Gab AI a description of the image to let it generate alternatives and it generated the following which all (more or less, probabilistic) fulfill the criteria:
That makes generating images on demand difficult. The same prompt can be fulfilled by countless images, esp. with the probability involved. And yup, it would be possible to make the prompt more detailed, but in the end, there are different versions possible.
So, great for getting different ideas for poses or compositions, but bad if you have a very specific image in mind.
Still an incredible technology, but at the moment, one prone to both artifacts and happy accidents.
Happy generating.
