Prompt: an art-deco style poster of a smiling lemming with the text "Welcome to Lemmy!" Below it. --ar 3:5 --v 6.0
I’ve been waiting eagerly for this update, looking in envy at those bing and chat-gpt images! :)
Not so much for the text, but for the greater coherence in the image and better understanding of the prompt.
Apparently the prompting is completely different and you’re expected to use more natural language. I’ll be happily experimenting for the coming days :)
Edit:
A few more examples. From what I can see the --stylize parameter has quite a bit of influence on how accurate and integrated with the image the text will be.
--stylize 50
--stylize 200
--stylize 350
--stylize 750
At really high levels the text is almost always accurate, but contained to its own little text box. Also it ruins the art style I intended to go for and it looks more like Fievel than a lemming :)
They’re busy with developing the ability to generate in their website, but for now it’s in preview for users with 10,000 or more images generated. I’m a heavy user, but not that heavy :) I think I’ll make a post if the website becomes freely available.
As for paying, I think they sometimes have free sessions but they’re very limited.
I'm in the 10k club and have played with it, it's pretty solid!