Someone actually created video content using this? lolol

They had to have access to a dedicated system; no way they went to the site generator to make these videos.

Didn't google make it open to all? The code or the AI or whatever make it work?

bobzchemist: i just love what deep dream did to C3PO ! and you can always get more absurd results by refeeding the image in the generator.
issabella: i really like that first one with the tree and the last one with the bigger creatures - what was the original picture of?
2-ton: cool find! i got rather dizzy watching them though .

are we getting into techie discussions in here again?

Thank you AnMal, it is always surprising each new result. I think that it was this one (I have another one very similar) the pic I used.

Anmal, that was 9 times through. It doesn't look like it changed at all between 7, 8, and 9. Am I doing something wrong?

issabella: that one looks right, considering the shape and colour fo the creatures . it turned out very nice when you deep dreamed it, i think i'll try a similar picture myself!
bobzchemist: i can't think of anything you are doing wrong, but maybe there's a temporary hiccup at the deep dream site? i compare results a lot when i refeed and i can always see differences between them, but sometimes the differences aren't very big. if there was no difference at all it seems the generator wasn't doing what it should, namely look for information in the image, interpret and enhance it.

I think I figured it out - to see noticeable changes, you have to change the image at least a little yourself before you upload it again. Here I started with a simple weavesilk image, fed it to deep dream, and brought it into GIMP. I duplicated the layer, added an alpha channel from the layer menu, and ran Color to Alpha on black. This gave me a layer without black. Just one layer doesn't make much of a difference, but if you duplicate it enough times (not the background layer, just the transparent one) the dark details from deep dream become much more vivid.

Original:


Deep dream:


Enhanced:

Second pass through deep dream:


Enhanced again:

the videos are so trippy to see.

bobzchemist: interesting with your experiments. you really got a high amount of detail on very few refeeds! i think the process could be simplified a lot by just adjusting saturation and contrast after the first deep dream feed and running "freaky details" or another detail filter in g'mic after the last.

Max Ernst did it without a computer!

https://ixquick-proxy.com/do/spg/show_picture.pl?l=english&rais=1&oiu=http%3A%2F%2F2.bp.blogspot.com%2F-xp84cDwpSVM%2FUPkvskQ8qtI%2FAAAAAAAAAYg%2FruWa7TWFd0w%2Fs1600%2Fla%2Bjoie%2Bde%2Bvivre.jpg&sp=d539922647fa6b777c887e7a0a45c263

erisian: painters are such cheaters, aren't they? instead of doing their research to find the best neural network image generator and the best way of getting what you want out of that generator and then put down hours of post processing the result in their image editing software of choice until they get something good, they just take their paintbrush and paint in the surreal details where they want them. i call that cheating!

Did it again. oops posted in the wrong place

sallyanne: wow i like that! really cool creatures and the hibiskus got such a dreamy, fuzzy look.

Thank you Anmal

sallyanne: that's beautiful!!! nice use of the dreamscope generator .

This is one of a purple flower carpeted street

to this

interesting how many people like the outcomes ot this "deep dream generator" (and how many nice results can come out sometimes, -the one before this last one for instance-).
For me they are not dreams but nightmares!....
but this is certainly me (16 pages of experiments demonstrate it)

I like the outcomes from deep dream generator and also from dream scope

Sallyanne - good job you got your computer working so that you can dream on

A result from multiple submissions until I got tired; may have to try Tran's automated method ...