As with my little Elevenlabs experiment, I've periodically played around with image generation models, mostly for producing character or setting concept shots. The results have been mixed. Text -> Image is a very awkward modality, even with the most recent editing tools. I hope one day we get something more like a 'real' image editor that seamlessly integrates traditional pixel-pushing tools with smarter generative components that can treat arbitrarily segmented parts of an image, as opposed to the current process which is a cross between 'found object' art and a slot-machine.

Stepping down from my soap-box, here are some of the better results I've managed to get. This first grouping covers a RepliFemme dealership and the robots that populate it.

These were generated as character concept images for a fembot in a WIP story.

And finally, these were another character concept, for a retro-SF WIP.

Nice art so far!

That's really great stuff! I want to see more of that retro-scifi setting.