Gemini 2.0 Flash Native Image Generation
# | #ai, #image-gen, #gemini
Gemini Flash Experimental now has native image generation capabilities - announcement.
I tried out a few things:
This one was oneshot.
Moving the woman to the right was not easy. It took me a like 10 tries to get this working.
This one required ~2 prompts for each image.
It's certainly rough around the edges and is difficult to get working. The prompting is weird and does not work as expected. I am not sure how temperature affects thigns. Regeneration leads to essentially same results a bunch of times - idk what that means.
It's insane that you can edit images now by talking to a model. It fast - 10s per generation. I expect it to get much better over the next few months and competitors (not Anthropic lol) to come out with similar capabilities as well.