Luminous Landscape Forum
Raw & Post Processing, Printing => Digital Image Processing => Topic started by: plugsnpixels on June 01, 2023, 12:17:30 am
-
I just got home from work. It's now dark out and it was a dreary SoCal day anyway (May gray/June gloom), no good for photos.
So I fire up the Photoshop beta on a 2020 M1 MacBook Pro and, working from a blank white canvas, create some fantasy Luminous Landscapes in minutes using Generative Fill...
Let's use this thread to post your own AI landscape renders. You can use the PS beta or one of the diffusion apps (even run them at the same time for efficiency of output), but let's stick to landscapes.
Here are my latest "babbling brook in the woods" renders.
-
I like these. I've never tried creating something using AI from a blank white canvas. Might be an interesting experience. Do you use any real-life pictures as references?
-
Step into the propeller...
Peter
-
I like these. I've never tried creating something using AI from a blank white canvas. Might be an interesting experience. Do you use any real-life pictures as references?
Thanks! I only used text prompts on the white canvas (approx. 8x12" @150ppi), then I use the result with an expanded canvas (maybe 20x12") to generate once again. This tends to give the wide-angle look.
I'll need to try generating from the final size canvas alone to see what happens.
I just tried expanding the canvas on one of my own post-processed (actual) photos, and here is the result. Obviously this adds potential to our existing work.
-
Step into the propeller...
Peter
I stepped into it right here... I was trying to do some nice beach cliff scene (PS only), then I added a pirate ship, then of course some pirates came ashore for a picnic!
Then I tried the same prompt in DiffusionBee...
-
I stepped into it right here... I was trying to do some nice beach cliff scene (PS only), then I added a pirate ship, then of course some pirates came ashore for a picnic!
Then I tried the same prompt in DiffusionBee...
WOW,You sure did...
:P
Peter
-
Following up in PS only using a larger canvas size to start. It's basically a larger version of the smaller renders (ie, same "focal length").
Adding the additional canvas and re-GF'ing gives the better (wider angle) results.
-
Grand Canyon (PS only), before and after (in the after I expanded the canvas via GF and used GF via Magic Wand selection to replace the river), then with DiffusionBee+PS (+post).
-
Blank canvas in Photoshop, prompt "luminous landscape", no further canvas enlargement. Clicked repeatedly for variations.
-
Back in 1985 I was a poor college student on tour in Europe and took black-and-white scenics up in the Swiss Alps-! Here is one of those (original + PS colorized, when doing that was cool on its own!), plus the recreation using PS-only GF from blank canvas (prompt: "winding dirt footpath in the swiss alps, green grass, wildflowers, snow on mountaintops, wooden fence, European barns, puffy clouds in blue sky").
-
Alps scene this time with DiffusionBee, then extended in PS with a bit of mountain and fence fix.
-
If I understand correctly Photoshop AI beta works by drawing on a library of images. If so, I look forward to seeing parts of my photos appearing in its creations.
;D ;D ;D
-
Far as I understand, it draws from selected Adobe stock, royalty-free and open source images. At this point the only way I know to purposely include our own work is to expand the canvas of our own images. Otherwise Adobe may never find us ;-)
-
And, if they used your image, along with other images you probably wouldn't recognize your 'information' buried in the soup.
-
And, if they used your image, along with other images you probably wouldn't recognize your 'information' buried in the soup.
Soup! That's what this is, yes! ;-) Adobe Chunky...
-
Begun in DiffusionBee, finished up in PS.
-
Some farmland somewhere, done in PS (text prompt in white image), extended with GF with elements added via GF, artsied up with Dynamic Auto Painter (https://order.mycommerce.com/cart/add?vendorid=200277176&PRODUCT%5B301022334%5D=1&affiliateid=200019699) (on M1 Mac in wine-preloader mode).
-
Starting with DiffusionBee and finishing with PS/DF (including massaging some elements) and more artsy.
-
photoshop renders photos meaningless...
-
Remember Thomas Kinkade? He's back. Check out this link.
https://briantubbs.medium.com/i-tried-generating-ai-images-in-the-style-of-thomas-kinkade-and-the-results-were-amazing-b671f6b1d570
Peter
-
photoshop renders photos meaningless...
1850s: Photography renders paintings meaningless ;-)
Crazy times, we need to roll with it.
-
Remember Thomas Kinkade? He's back. Check out this link.
https://briantubbs.medium.com/i-tried-generating-ai-images-in-the-style-of-thomas-kinkade-and-the-results-were-amazing-b671f6b1d570
Peter
Great article, thanks!
-
Well, the macOS Sonoma beta on M1 MBP broke DiffusionBee and Draw Things (both crash while rendering...). Exploring other options.
-
Trying the free "Amazing AI" app from the Apple App Store to work around DiffusionBee (etc.) crashes in Sonoma. Bangs out 4 results at a time, nice big 2048 × 2048 renders but soft, needs artsy effect added to cover up (not applied yet here).
Performed my usual PS GF canvas expansion trick with today's PS beta update, seems much faster.
-
A quick trip on the river, done in PS alone.
Remember! Take only photographs, leave only footprints...
-
1850s: Photography renders paintings meaningless ;-)
Crazy times, we need to roll with it.
Not really the same. Photographers weren’t pretending that their photographs were paintings. However, going forward, we aren’t going to have any idea what is an actual photograph and what is an AI composite. Don’t know about you but that difference is very important to me. And I can’t wait for the 2024 Elections to see all the photographic evidence of all the awful things Trump and Biden have done. It’s a photograph, must be true, right?
-
I hear you! To be specific, in my own work I am focusing on harmless AI scenics (which I then like to turn back into faux-artwork, ironically enough), but I do agree we have entered a place where "nothing is real", even worse than usual. Same with audio, now we have voice models being applied to songs the original singers never made (the best use I've heard so far is doing something about Paul McCartney's current voice...).
Didn't we recently have a situation where a fake (possibly AI) Pentagon attack photo (as I recall, some smoke and a building in the background that didn't even look like the Pentagon) triggered a stock market decline?
Defintely going to be interesting. So I will stick with luminous landscapes, can't go wrong there...
-
Got DiffusionBee working again on my M2 Mac Mini.
Followed up as usual with post in PS GF.
Life on the farm is kinda laid back...
-
Got back to it last evening, DiffusionBee + PS beta GF.
-
Now Adobe adds Generative Expand, where you use the crop tool to quickly do what I was manually doing earlier with increasing canvas size and making selections.