site banner

Small-Scale Question Sunday for October 8, 2023

Do you have a dumb question that you're kind of embarrassed to ask in the main thread? Is there something you're just not sure about?

This is your opportunity to ask questions. No question too simple or too silly.

Culture war topics are accepted, and proposals for a better intro post are appreciated.

2
Jump in the discussion.

No email address required.

Does anyone have access to ChatGPT with images? Is there some opt-in setting I have to find and enable, or are they still doing a staged roll-out?

I have Dalle3 accecss. It's... alright.

Humans look extremely uncanny. All pictures look like 3-d renders.

Worst of all, They reword your prompt in 4 different styles.. why? It's even funnier when you want pictures of people. It will automatically give you 4 different races (evident in the prompt lol).

I'm really not liking how much OpenAI tries to baby its users.

We using the same DALL-E 3?

Because for me, it's amazing, almost solving the problem of AI image generation barring truly baroque prompts.

It can be as aesthetic as Midjourney, while having far more intelligence, especially for semantics and interactions between objects, than either that or SD.

Previously, asking for artwork of a woman with cyborg legs holding her child while they curiously look up at a painting of a superhero would have produced less than ideal results, like a female cyborg child looking at her mom or something. It usually works these days.

And humans look human, in every art style I cared to try, which is an improvement from the previous DALL-E 2 Experimental, itself far better than the first DALL-E 2.

Here, pictures are a thousand words or something along those lines, depending on the compression algorithm I bet:

And I'm cherry picking the best out of 2 or 3 acceptable images in a batch of 4, not one from 4 batches like I had to previously