Thanks, but I do this on Android and there is no little circle-i on the images (or any equivalent way to access the type of data I’m talking about). I also don’t go through the chat interface to generate images but create them directly.
Are you using the ChatGPT android app, or a particular mobile browser to get to the web interface? I’ve tried both, but not often. ISTR being able to get to the image prompt on both, but I might be misremembering.
Also please explain what you meant by not going through the chat interface?
Leonardo.ai’s Realtime Gen is getting pretty good. It allows real time generations without using up credits, then you can choose to upscale (for credits), if you want. Not quite as photo-realistic as Midjourney, but getting there.
Here’s a Leonardo MASH rendering after I face-swapped using InSwapper:
I mean this:
No interaction with a chatbot, so no asking it what a previous prompt was.
Ah, when you mentioned GPT in your post I assumed you were using ChatGPT as the intermediary. I’m not sure there is any kind of intermediary when using it via bing, but I haven’t investigated that UI. I’ve been happy to let ChatGPT-4 and its derivative specially trained GPTs do the magic for me.
There very definitely is. I’ve posted multiple examples in this thread (including the one you responded to). Search the thread for “ethnically ambiguous”, for example.
I love how right it gets the SoCal mountains.
Here they are extended to 16:9 with Adobe Firefly generative expand:
-love the shorts it put on Pierce and Hawkeye.
Pierce and Hawkeye are the same person, but further, it’s totally inaccurate because Trapper was taller than Hawkeye.
Brain fart. And I did remember that Trapper was taller than Hawkeye, but the face-swaps were better with them the other way.
Bing, show me Hot Lips and BJ.
Bing:
The priest should not have let Regan go out for a bike ride.
I wanted a front view of a girl on a bike but mentioning she had a ponytail flapping in the breeze tended to get things like that.
Stability.ai has a new image model:
I watched a youtube video of some of the putput, and it’s pretty good.
Did it get the ball through the windmill?
The game of putput does not use a ball. You should know that. The windmill, however, figures prominently. Especially in the second quarter.
Here’s the model.
It produces good images but doesn’t understand complex prompts well. And it can produce only very brief text.
It is interesting that you see the image in progress.
No celebrity filters yet. It was confused by Bob Ross painting Taylor Swift riding a pink unicorn but did a reasonable job when I dropped the unicorn.
That Taylor Swift doesn’t need a filter 
Bob was working outside his bailiwick.
Text to video is improving a bit.