next »

[gouhadouken]

Convincing skin tones! pseudo code:

float r = random.range(0.05f , 1.0f);

float g =  (r * 0.8f) + random.range(-0.05f , 0.05f);

float b = r * 0.5f;

Color skinTone = new color(r , g , b);

[pelle]

Convincing skin tones! pseudo code:

float r = random.range(0.05f , 1.0f);

float g =  (r * 0.8f) + random.range(-0.05f , 0.05f);

float b = r * 0.5f;

Color skinTone = new color(r , g , b);

I turned it into a python script while drinking my morning coffee and it looks good, except I think (on this monitor anyway) some of the darker tones have a bit much green and/or blue in them (although my knowledge of the full range of human skintones might be lacking?). I do not know the reasoning behind the exact values used, but maybe the range for g (and b?) need to be scaled to decrease more for smaller values of r?

Code:

import random
import sys

nrtones = int(sys.argv[1])

print "GIMP Palette"
print "Name: gouhadouken skintones from tigsource"
print "#"

for n in range(nrtones):
    rf = random.uniform(0.05, 1.0)
    gf = (rf * 0.8) + random.uniform(-0.05, 0.05)
    bf = rf * 0.5
    r = int(255 * rf)
    g = int(255 * gf)
    b = int(255 * bf)
print "%d %d %d\t#%02x%02x%02x" % (r, g, b, r, g, b)

[BomberTREE]

@Prinsessa
I enjoyed playing your game, your little generated worlds are fun and the audio is 

Could you guys post an image of the skintones?

[Zireael]

I can procedurally generate a car (the one at the side) and some wheels, and the road, and the nameplate and the label. The buildings are just procedurally placed, not generated yet. All the things I mentioned are procedurally generated in engine, not imported from Blender or w/e.

[oahda]

@Prinsessa
I enjoyed playing your game, your little generated worlds are fun and the audio is 

Thanks from both of us! <3

[gouhadouken]

@BomberTREE

W/o rng:



W/ rng:

[gimymblert]

@BomberTREE
W/o rng:

Industry standard practice

W/ rng:

Cutting edge algorithm


Black skin a bit greenish though

Remind me of this:
http://www.mediainstitute.edu/media-schools-blog/2013/12/the-vector-scope-lessons-of-the-flesh-tone-line/
http://www.lafcpug.org/tutorials/basic_fleshscope.html

[Artsicle]

Try doing a bigger sample size.
If you get these results again and again maybe add a higher chance for white

[gouhadouken]

Try doing a bigger sample size.
If you get these results again and again maybe add a higher chance for white

just based on what I've seen so far, the euro look only comes when the red channel rolls > like 85 - 90%, so there's that. There's defs room for tuning in the gen as well as the distribution of results. This basically just came about from me color picking fleshy tones in PS and eyeballing the rgb sliders.

Also, @BomberTREE those links are mad useful and your comments made me lol

[gimymblert]

Also, @BomberTREE those links are mad useful and your comments made me lol

Wait are sure it's bombertree? I was the one posting links lol

[jgrams]

Yeah, those links are neat.  I never realized that.  Here's a try with HSV: hue from about 14-25 degrees, saturation from 0.4 to 0.55, value from 0.5 to 1.0.

skin-tones.love

It's interesting how much natural variation in hue there is within the face: these all look kind of flat and unnatural because they don't get more and less red with the angle of the light.

[gimymblert]

Documenting surface variation of the melanin is the most difficult part of my current research, at least for black people ... everybody want me to make black cheek "reder", even though I show photo of people where it's turning toward orange, and pointing to all sort of subtle variation they straight up don't see (subsurface don't seem to behave the same). Then there is stuff I perceive but don't see ... and there is reference material where people are covered with make up, so bad reference ... Skin is complex.

If some of you want to dig deep into this I recommend at least those paper:

A Spectral BSSRDF for Shading Human Skin by Craig Donner and Henrik Wann Jensen
Which basically about generating base skin color accurately depending on the various parameter (blood oxygenation, melanin mix, etc) but using integration over the XYZ lighting model

A Layered, Heterogeneous Reflectance Model for Acquiring and Rendering Human Skin
Talk more in depth of the various layer and variation on the skin that vary colors like blood vessels etc ..

[gouhadouken]

Also, @BomberTREE those links are mad useful and your comments made me lol

Wait are sure it's bombertree? I was the one posting links lol

I'm a special boy :/

[Josh Bossie]

So, I'm trying to bring some concepts of procedural generation to my game, Village Monsters

It's a life sim game, and because I'm passionate about simulating systems I've included a lot of them - simulated time, weather, creature behavior, etc.

As I developed these systems, I realized that I wanted to do more than just randomly pick a weather type for a day, randomly spawn critters depending on the season, etc.

Eventually I came up with the idea of a simulation controller I'm calling the Director - in truth, this name (and concept!) is very much stolen from Left 4 Dead.

The idea is that the Director has access to all the simulated systems and uses this information to control when things are created, how often, and more.

The reason I'm going with Director for the name is that each (in-game) day he picks a personality and 'directs' the day based on it. Some days he might be feeling generous and decide to purposefully spawn bugs and fish and furniture the player hasn't found yet. Other days he might be feeling a lot less generous and give you very little.

The eventual goal is to allow the Director to create entire patterns to give days or weeks a certain 'feel'. For example, he might be feeling "gloomy" and decide to make the entire week full of rain storms, darker-than-usual lighting, and somber ambient music.

It's also very experimental still, but I'm having a lot of fun applying "procedural" concepts to something beyond terrain for once

[pelle]

Chris Crawford describes something that sounds like that Director in his book about interactive storytelling. Used to keep an eye on the different subsystems and make sure that good stories are generated instead of just random randomness, but I do not remember what he called it. Also not sure if he ever managed to actually implement it or if it is all just theory, but it sounds like a good thing, and maybe can work at least for reasonably simple systems.

[Pishtaco]

I'm trying to generate some continents, one or two thousand km on a side, for a top-down flight sim. The current system generates a plausible original shape using perlin noise, adapted to give some discontinuities (faults and rifts) and some remains of old mountains. Then I run several million years of fluvial erosion and mountain growth.

You can try the current version: https://dood.al/test/surface_transport.html

[gimymblert]

I hope someone found a way to have erosion like feature without simulation one day ...

[Josh Bossie]

Chris Crawford describes something that sounds like that Director in his book about interactive storytelling. Used to keep an eye on the different subsystems and make sure that good stories are generated instead of just random randomness, but I do not remember what he called it. Also not sure if he ever managed to actually implement it or if it is all just theory, but it sounds like a good thing, and maybe can work at least for reasonably simple systems.

I hadn't heard of this before, but I'll take a look - that's right up my alley. Thanks!

[TonyLi]

Has no one mentioned procedural quest generation yet? Bethesda dabbled in it with their Radiant Quest System. Privateer even implemented a basic but wholly fun mission generator back in the day; I could always hit up a terminal for something entertaining to do.

I'm not posting to promote the procedural quest generator I just released for the Unity Asset Store (you can read about it here if you want) but since the thread is moving away from map generation into narrative stuff like L4D's Director, I thought it might be interesting to share a bit of info. My main influences, apart from classical AI planning, were Till Reimer's Agent-based procedural quest generation for a role-playing game, Bryson & Grey's Procedural Quests paper, and a bunch - of - papers by Doran & Parberry. And a ton of feedback from folks on the Unity forum over the last 2 years.

One challenge was figuring out what info to annotate the game world with to give it a rich variety of possibilities without burdening the designer too much. In the end, I ended up with entities that have actions, and quest givers that are aware of entities in designated areas. To generate a quest, the quest giver chooses an entity that it deems urgent, and then it plans a sequence of actions to improve the game world from its point of view. It models factions and personality values, so different quest givers could come up with different quests. For example, say there's a sick villager entity. A doctor NPC might generate a quest to heal the villager, while a thief might generate a quest to rob the villager and split the loot.

There's quite a bit more to it, but I think procedural quest and narrative generation has so much untapped potential. There's an opportunity right now for indie devs to really stand out by generating this kind of content in an interesting way.

[gimymblert]

Hi TonyLi welcome to Tigs Board, I'll soon buy and experiment with your asset and implement idea I had! Continue the good work!