The background for these posts can be found in THIS post.
In brief, these posts serve to introduce new readers — and reintroduce regular readers — to photos from the early days of this blog and, occasionally, to photos from days before this blog came into existence.
Today, I offer a gallery with 392 photos originally shared in a series of posts (HERE), but if you want a good introduction, look in THIS post. Sorry, I should mention I’m talking about DeepDream Generator.
You can read all about it, and you can probably get lost playing with it, but the gist of it is this: making use of Google’s open-source DeepDream code, the site allows you to merge two photos to create a new one. That’s the simplistic version and you can replace “merge” with overlay, marry, join, or any descriptor you want that elicits the idea of taking textures and details from each photo and compiling them into a new photo.
It doesn’t have to be two photos; it could also be one photo and adding textures and/or styles to them (a bit like what I do with Topaz Studio and Impression, but more). However, I found that using two of my photos — judiciously picked — I could get some nifty results.
That’s a small photo, and it’s the actual size. Let me explain.
The site is free to use, but you ‘earn’ functions based on your level of participation. Meaning, that when you first start, the resulting photos are limited in size, and you may be limited in both the speed and options for the rendering. Then, as you keep using it, you earn credits and can output larger photos, render them faster (more resources allocated), and you have a few more options as to how the photos are compiled.
Some of what I could produce mimicked what I could do in Impression . . .
Note the appearance of eyes and/or dogs in the photo. That’s a common thing when using only one photo and some of the canned processes. I don’t find it attractive, but some people do.
One of the explanations I’ve read is that the program samples photos on the Internet and because there are a lot of dog photos and photos with eyes, we get the above result. I don’t know this to be valid, but based on how the program works, it makes sense.
Regardless of the actual mechanism, once you get the feel for how things will integrate from one photo into the next, you can judiciously choose textures to get the effect you want.
I mean, yes, there’s some trial and error aspect to it (experimentation) but if you find one photo that creates interesting elements, you tend to use it on other photos.
Photos of lava rocks worked well for some subjects . . . .
Bagel faces and seashells on a beach also proved interesting . . .
Some of my favorite favorites came about with photos enhanced by the patterns on an old Singer sewing machine . . .
Even after you ‘earn’ enough participation to be able to run larger files, they are slower (more computing power) so I would often still run smaller files to see how they do, so I have a mix of small and large files even after I’d earned enough credits to do larger offerings.
. . . of course, the gallery has a fairly decent mix of “looks”, so if interested in exploring more, you can find a good variety of treatments . . .
OK, just two more; same subject, different looks . . .
. . . but, no worries; plenty more in the gallery.
Note: the transition is set to 2sec, but — if you move the cursor anywhere within the photo — you’ll see a pause button on the lower left, and, once paused, you can use the left and right arrows on both sides of the photo to navigate the slideshow. If you click anywhere in the photo instead of the pause button, you’ll exit the slideshow and find yourself in SmugMug. You can still scroll through the photos, or interact in other ways.
That’s it. This post has ended . . . except for the stuff below.
Note: if you are not reading this blog post at DisperserTracks.com, know that it’s copied without permission, and likely is being used by someone with nefarious intentions, like attracting you to a malware-infested website. Could be they also torture small mammals.
Note 2: it’s perfectly OK to share a link that points back here.