They have an AI engine that you can type what ever you want and it will paint the interpretation of your command.

This is AI, and supposedly completely original, painting attempts with a command of “Pontiac Fiero”.







I asked to paint a Fiero in space.

[This message has been edited by Wichita (edited 06-12-2022).]

Tell it to stop looking at wonky kit cars and try draw a fiero.

From a programming / AI / neural network aspect, this is really cool.

They look "wonky" because the AI has probably found tens of thousands of pictures tagged as "Fiero", and it's trying to understand and piece how all these different shapes, angles, views, and mods can all be the same thing. The technology is more advanced with linguistics, even in a language as difficult as English. They have been able to use the visual approach to build AIs that can look at photographs and "find" specific objects, such as cats.

So this next step of saying "not only learn how to identify this thing, in all it's variations, but build that into an original concept and draw/paint it". These might look wonky, but it's not much different than the effective drawing level of any toddler. Pretty awesome stuff technologically.

</end tech geek>

This is a lot of fun. I just discovered Dall-E today so naturally my first try involved a Fiero. This is “Fiero car in the movie Tron”:



This is “1967 Pontiac GTO as an army tank”:



You can try here:

https://www.craiyon.com/

[This message has been edited by Quadfather (edited 07-09-2022).]

The version of Dall-E you are using is the public version which can do perhaps 1% of what the "true" version can do. I've seen many samples of the full version and it's absolutely mind-blowing.

quote Originally posted by Cliff Pennock: The version of Dall-E you are using is the public version which can do perhaps 1% of what the "true" version can do. I've seen many samples of the full version and it's absolutely mind-blowing.

Very few have access to Dall-E2 full version, but you are correct. It is stunning.


DALL-E 2 created this image in response to the text "teddy bears mixing sparkling chemicals as mad scientists in a steampunk style"



"a photo of a quaint flower shop storefront with a pastel green and clean white facade"



"a bowl of soup that looks like a monster knitted out of wool"

[This message has been edited by Wichita (edited 07-10-2022).]

"Spider-Man From Ancient Rome"

There are plenty of really interesting YouTube videos on this technology. It is truly mind blowing. What surprised me the most was how long it takes for the AI to create these intricate masterpieces. Not several hours. Not ten minutes. It takes just twenty seconds per image!!!

I like the following video, as not only does it show some perfect examples... it also shows some less than perfect examples... and how to fix them.




The next step for this fantastic technology will be to advance from still images to moving video and/or to 3D holograms.

[This message has been edited by Patrick (edited 07-10-2022).]

quote Originally posted by Patrick: The next step for this fantastic technology will be to advance from still images to moving video and/or to 3D holograms.

Video is basically just progressive images, i don't imagine it will take long at all. And we can already deepfake voice for adding audio. I imagine it is completely feasible to have an AI generate sound effects based on in frame imagery.

Holograms are still a ways off i think, but this tech could work wonders for 3d rendering. imagine being able to tell it you want a model of a specific car, and it could just make you one!