Previous Entry Share Next Entry
breeden
ursulav

Fun with Filters

You know, back when I was a budding digital artist, filters were the devil. You could always tell them and they looked so cheap. But the times, they are a-changing...

So I downloaded the Prisma app for my phone and have been having far too much fun these last two days with it. Mostly I just can't get over how much the tech has improved in just...uh...

Okay, on second thought, let's not do the math.



I always wanted to live in a Studio Ghibli painting...


...or a Hokusai. Though not the wave.



Now I want to do a photo comic.



No, no, I don't have time. The workflow alone of the photos to photoshop, then to phone to filter, (there's no desktop version that I've found) then back to Photoshop would make me bonkers and as with all artistic shortcuts, would not actually save any time. But I can totally see this taxidermied boar head from the Siler City antique barn in a story. Possibly as a quest giver.

The kudu is skeptical.



(The other problem is that these only output in 1024 x 1024. They'd be much better for a weird little Ren'py game...NO! BAD ARTIST! Too many projects already!...still...NO. The art is the fast part! The writing and the coding is the hard part! Brain, stop!)

Finally I pulled up an old unfinished painting of a corgi warrior (original going off to live with the player of the thief from our D&D game) and had some fun with it:



Or after Hokusai:



Soon enough I suppose we'll be able to ID all these filters easily--and lord knows, you still need a very particular sort of photo, you can't just filter anything, so it's not like you can just turn your vacation photos into a comic--but I'm having far too much fun with it as it stands.

Already I'm trying to figure out how to do the Lichtenstein style manually by breaking an image into flat layers and pasting the patterns in. Which shouldn't be too hard, in theory...

Probably other artists work on art by having something to say, but these days, I am largely out of interesting things to say and so mostly proceed by going "How would I do THAT? Okay, let's figure it out!" As a wise man once said, a technical challenge may not be as good as a creative one, but it'll do in a pinch.

Anyway, the app is called Prisma, it's free, it's fun, it works best with very crisp images with good color separation (you can see where similar values get murky in the frog's back foot, for example) but it's fun to play around with.

  • 1

Fun and discovery! Woot!

are those the Queens Own Corgie's or a different unit?

I installed that last night thanks to your tweets. I don't know if I should really thank you for that. XD

Those ARE wonderful filters; and as you say, you do need the RIGHT starting image or it's all a mess.

These sorts of programs work better when they have actual information about the geometry rather than having to guess from the picture colors - so IE with a data feed coming in from a CGI program rather than flat photographs. But that's a whole other workflow . . .
I await the day when my phone has two or three cameras to catch some real depth information to feed into these things.

I love what you have done with the photographs. They really do look like paintings. I would love to have a Totoro at the bottom of my garden. That boar's head is almost nightmarish, I can't help wondering if the original photograph was better or worse?

The boars head is definitely my favorite and it would be such a perfect quest giver!

I have a friend who's been posting some gorgeous prisma art on Facebook, so I finally gave in and got the app last week, but have yet to try it. I prefer to edit photos on an actual computer, and moving files back and forth is more work than I want to do most days, but the results are stunning.

I love both versions of your corgi painting!

There are android emulators you can install on your desktop that will allow you to run phone apps on your computer. Might help with the workflow issue.

My Prisma app on my iPad will often lose the save/export/etc. commands after I save and/or export one picture, requiring a restart of the app (double-click home, close the app in question). But it is still SO COOL.

I blame you for the sudden exponential rise in numbers of pictures in my photos app... ^_^


I have trouble imagining a corgi choosing to have piercings, after my parents' corgi mix, Cody. He was a colossal wimp when it came to pain or even threat of pain. If you bonked him with an empty soda bottle and it made that 'PONG" noise they do, he'd run away with high pitched 'Yi-Yi-Yi-Yi!' yelping you usually associate with injury in dogs. But he was just scared by the sound it made when it hit him.

He got his ear torn when a neighbor dog attacked him; it stayed rather tattered because my parents' didn't see a need to get it stitched, especially since it's cartilage so it's hard to stitch and likely to scar badly if it even heals back together. Well, when my fiance and I moved across the country, I flew out a week in advance and stayed with my mom while my dad and my fiance drove the car and brought his emotional support dog (formerly our pet dog), Bailey. Cody was a bully, and he tried to bully Bailey after 6 days in the car driving cross country, and she was just NOT HAVING IT. She managed to bite the tattered ear, which bled a bit and sent him running off yelping until he made it to his bed.

So I started searching on how Prisma works, and apparently, the resulting image isn't generated on your phone -- the source image is sent to the cloud with a server running a neural network with a GPU, and the neural network creates a new image based on your source with the style you chose, and then sends the resulting image back to your phone. Whoof.

So there's already a network round trip involved.

Update: Newer versions of Prisma have some filters that work in offline mode; ie, there is no roundtrip to the cloud for those specific ones. See the Wikipedia page for "List of Prisma (app) filters" for information on which filters specifically.

Anyway, it turns out that there is a website that claims to do something similar!

deepart DOT io SLASH hire

I haven't tested it at all, but I am curious to know if it works sufficiently similar to the Prisma app to cut out the step of sending the source image to your phone.

(edit to add update)

Edited at 2016-10-17 07:27 pm (UTC)

If you join the notaspammer community, your account will be added to a whitelist of users who are exempt from spam protection restrictions. You'll be able to comment with regular links without having to mangle them.

Edited at 2016-10-18 08:23 pm (UTC)

Deepart allows you to choose your own style, as well, which means that you can swap the source image and style image -- so in addition to creating "Corgi Warrior in the style of Hokusai's The Great Wave Off Kanagawa", you could create "The Great Wave Off Kanagawa, in the style of Vernon's Corgi Warrior".

Heh.

And now I want to see that!

I couldn't find the original corgi warrior, so instead I tried "The Great Wave Off Kanagawa, in the style of Vernon's The Biting Pear of Salamanca."
9b34b84e44814b7a86c72bf6a9129d77.jpg

That wasn't so great, so next I tried "The Great Wave Off Kanagawa, in the style of Vernon's Sir Bunny vs. the Wockwurm."
faecdd1853f59bf80c1be968800c76a3.jpg

I like that one much better.


Edited at 2016-10-19 02:18 am (UTC)

Awesome! and I too like the second one much more than the first. much thanks for indulging me in this.

(Deleted comment)
Late last year I was toying around with a Twitter bot that did something similar called DeepForger. A lot of the results were interesting but kind of boring, but my shadow frog icon redone in the style of Andy Goldsworthy was kind of neat.

I recall reading a Twitter conversation between the bot's author, another AI researcher, and a traditional artist that eventually came around to the idea that sometime in the future an artist might create "style images" that inform AIs like this one on how to create artwork in that artist's style. So if you commissioned an artist to create something they might feed a basic sketch or layout into the AI, along with their style image, and out will pop the finished product.

Edited at 2016-10-18 09:28 pm (UTC)

I've been wondering how to get good images for a Ren'Py game. Running a filter over my own photos could be a great start! Thanks!

Your eye for these details is fantastic. I can't wait to see the art it inspires.

  • 1
?

Log in

No account? Create an account