AuthorTopic: Possible activity discussion.  (Read 154257 times)

Offline Atnas

  • Moderator
  • 0100
  • *
  • Posts: 1074
  • Karma: +2/-0
  • very daijōbs
    • paintbread
    • paintbread
    • View Profile

Re: Possible activity discussion.

Reply #130 on: November 17, 2011, 01:13:22 am
Just a bit of a quote from the off topic thread to redirect the discussion here:

I was wondering if anyone can help me find a program someone posted awhile ago on here. It generated images that looked like this:


I tried the search function but I wasn't able to find anything. I think Happymonster created it but I'm not sure.

Yep!
Found an old mirror of it on my site from 2009. It's called Richard's Evolving Sprite Tool, for reference.

Awesome little program, a challenge/activity involving it would be neat.

I would love to participate in some kind of a challenge involving it.

Second this notion. :y:

Those are wonderful, and a CHALLENGE! centered around this would be ace. Question is, what should the challenge be?

Brainstorm!

Two obvious notions that come to mind are using a randomly generated base and pixel what you think you see in it. or the other way around, turning something existing into one of those, which in my opinion is the unattractive cousin of notion #1.

Offline Atnas

  • Moderator
  • 0100
  • *
  • Posts: 1074
  • Karma: +2/-0
  • very daijōbs
    • paintbread
    • paintbread
    • View Profile

Re: Possible activity discussion.

Reply #131 on: November 17, 2011, 01:41:37 am
Here's the link to the program in case you missed it. Happymonster made it close to 3 years ago I think.

An example I made a few years back of the 1bit patterns interpreted as sprites:



Anyway, ideas:
  • It could be fun to have a reservoir of 1bit patterns that people can add to with interesting ones they've generated. Kind of freeform in that you could post 5 or so designs if you dont have time to do art and then others can take those and make sprites from them without having to look for cool patterns.
  • That might become troublesome to organize all the cool patterns everyone makes, so it might be better to create a topic where we gather together > 50 interesting designs and then have a separate challenge where you have to make at least 5 of them into sprites.
  • Another option is to just have the thread there for everybody to post their own 1bit sprites and their interpretations. While being the most straightforward approach (least management), it loses the fun of having a few designs everyone can interpret differently. Then again, people can just do what they want so this point is kinda useless.

Ideas? Probably January or February is when this would happen.

Offline surt

  • 0011
  • **
  • Posts: 570
  • Karma: +0/-0
  • Meat by-product
    • not_surt
    • http://pixeljoint.com/p/2254.htm
    • View Profile
    • Uninhabitant

Re: Possible activity discussion.

Reply #132 on: November 17, 2011, 11:28:07 am
I mocked up a super simple thing like this in HTML5/JS here.



It is the simplest it could possibly be just every pixel an even chance of being foreground or background and mirrored.
There's a wierd bug in Firefox where every sprite is identical (RNG seems to be getting reset or not reseeding, if I stick an alert in between tiles they do generate nicely, wierd. :crazy:), but it works fine in Chrome (apart from the ubiquitous blur).
Refresh to regenerate. If you've got chrome hit the image button to get an image to save.

Offline Atnas

  • Moderator
  • 0100
  • *
  • Posts: 1074
  • Karma: +2/-0
  • very daijōbs
    • paintbread
    • paintbread
    • View Profile

Re: Possible activity discussion.

Reply #133 on: November 17, 2011, 01:10:45 pm
Awesome! Any chance for further development? The original has the source visible iirc. If you have a problem with the web one I could port it to mac/linux.

Offline rikfuzz

  • 0010
  • *
  • Posts: 427
  • Karma: +1/-0
    • View Profile
    • twitter @hot_pengu

Re: Possible activity discussion.

Reply #134 on: November 17, 2011, 02:02:35 pm
I mocked up a super simple thing like this in HTML5/JS here.



It is the simplest it could possibly be just every pixel an even chance of being foreground or background and mirrored.
There's a wierd bug in Firefox where every sprite is identical (RNG seems to be getting reset or not reseeding, if I stick an alert in between tiles they do generate nicely, wierd. :crazy:), but it works fine in Chrome (apart from the ubiquitous blur).
Refresh to regenerate. If you've got chrome hit the image button to get an image to save.

Works fine for me (Firefox 8.0 OSX), really cool!  Be really interesting to tweak it. 

eg.
Closer to the edge slightly less chance of being on, closer to centre slightly less chance of being off might avoid so many looking so square.

Next pixel influenced by previous pixel (again very slight influence) to promote larger clusters and less noise. 

Offline surt

  • 0011
  • **
  • Posts: 570
  • Karma: +0/-0
  • Meat by-product
    • not_surt
    • http://pixeljoint.com/p/2254.htm
    • View Profile
    • Uninhabitant

Re: Possible activity discussion.

Reply #135 on: November 18, 2011, 12:05:52 am
Closer to the edge slightly less chance of being on, closer to centre slightly less chance of being off might avoid so many looking so square.
Next pixel influenced by previous pixel (again very slight influence) to promote larger clusters and less noise.  
Done and plan to do.

Updated:
You can now specify parameters using GET parameters in the URL. eg. http://tools.putpixel.com/spritegen.html?size=12&spacing=2&probMin=0&probMax=1&falloff=cosine
Falloff so fewer pixels at the edges. Falloff types: constant, linear, cosine, spherical. Defaults to linear.
Now randomly mirrors in x, y, both or neither.

Thing: Set spacing to 0, falloff to constant, probMax not too high and go hunting in a tapestry.
« Last Edit: November 18, 2011, 12:51:58 am by surt »

Offline Redshrike

  • 0001
  • *
  • Posts: 93
  • Karma: +0/-1
    • View Profile

Re: Possible activity discussion.

Reply #136 on: November 18, 2011, 01:50:21 am
I mocked up a super simple thing like this in HTML5/JS here.



It is the simplest it could possibly be just every pixel an even chance of being foreground or background and mirrored.
There's a wierd bug in Firefox where every sprite is identical (RNG seems to be getting reset or not reseeding, if I stick an alert in between tiles they do generate nicely, wierd. :crazy:), but it works fine in Chrome (apart from the ubiquitous blur).
Refresh to regenerate. If you've got chrome hit the image button to get an image to save.
I've spent quite a bit of time playing with RIST over the years.  I think your may produce better results, though of course the "evolve" button was fun.

Offline surt

  • 0011
  • **
  • Posts: 570
  • Karma: +0/-0
  • Meat by-product
    • not_surt
    • http://pixeljoint.com/p/2254.htm
    • View Profile
    • Uninhabitant

Re: Possible activity discussion.

Reply #137 on: November 18, 2011, 02:17:32 am
I've spent quite a bit of time playing with RIST over the years.  I think your may produce better results, though of course the "evolve" button was fun.
I should hope not given that mine works purely randomly. Most of the work is in the human brain anyways, to draw meaning out of the noise.

I've found a workaround (not a solution) for the Firefox bug in case it was affecting anyone other than me.
« Last Edit: November 18, 2011, 02:23:26 am by surt »

Offline Redshrike

  • 0001
  • *
  • Posts: 93
  • Karma: +0/-1
    • View Profile

Re: Possible activity discussion.

Reply #138 on: November 18, 2011, 04:49:10 am
Random, but not all randomness is equal.  Something about the proportion of black to blue, the spacing and the patterns of mirroring seems to give me better results in terms of "seeing things."

Offline surt

  • 0011
  • **
  • Posts: 570
  • Karma: +0/-0
  • Meat by-product
    • not_surt
    • http://pixeljoint.com/p/2254.htm
    • View Profile
    • Uninhabitant

Re: Possible activity discussion.

Reply #139 on: November 18, 2011, 10:04:13 am
Added despeckle: remove single pixels by removing the pixel or adding a neighbour (parameters: despeckle [off, remove, extend, random], despeckleProb); and despur: remove pixels with single neighbour (parameters: despup [off, on, random], despurProb). eg. http://tools.putpixel.com/spritegen.html?despeckle=remove&despeckleProb=1&despur=on&despurProb=1&size=24&zoom=2, http://tools.putpixel.com/spritegen.html?despeckle=remove&despeckleProb=1&despur=on&despurProb=1&size=12&zoom=2
« Last Edit: November 18, 2011, 10:44:38 am by surt »