Wednesday, October 15, 2014

Pixel tutorial

My ability to draw has always been hit or miss, I had fundamental classes in high school and would occasionally doodle, but always avoided the human figure because of its difficulty. And I still find it so, probably always will. But through doodling and what I retained from the classes I've taken so far in drawing I've picked up a few hints that helped.

1. Practice/mentality, it's a duh moment but it's true. Most important though is to practice the right way. When you're just drawing things off the top of your head and you're having trouble making them look photorealistic or "life like", then I suggest trying to draw something out of a photo. This isn't tracing, this is exercising the eyes' ability to notice and the hands' ability to translate that. Drawing is 99% in the head, and 1% in the hands and eyes. If you can see it, you can draw it as you see it. Where you start doesn't dictate where you end up. The trick is to exercise that link between the hands and eye. Seriously, anyone can do it. To quote Garfield from the animated series Garfield and Friends, "it's just lines on paper." And that is unarguably all it is, just like how paintings are just paint on canvas. And it's not in any explicit way of how that paint is put on, there isn't any technique or way to move that dictates how well the lines end up looking. It's in seeing what the lines do, and recognizing what that means in relation to the piece as being presented as you work, and keeping that in mind as you work. What this means is as simple as "that line on his ear comes in too far, makes his ear look too big".  And then making the line come in a little. That is literally all drawing is. A bunch of that one, line and alteration after the other until you feel happy with it. Reason experienced artists make it look so easy is they've exercised that hand/eye/brain coordination so much they can more quickly approximate what they see in their head on paper, where with the rest of us faces start off looking vaguely potato-ish, they know how to draw the right face shape from practice. Just like those pieces you see the tourists artists do in no time at all that come out looking good, they've practiced how to make what they see so often their hands just know what to do.

Now, I like pixel art for two reasons. First because regardless of how many mistakes I make the canvas doesn't show so I always feel in control of the finished product, and secondly because there is a total abandon within having definite color and space. I can pick this or that pixel to color, but no where in-between. I have a limited pallet and can only attempt to replicate a degree of what colors exist in real life. While, say, this image

has a massive number of shades of grey and very specific locations for the white flakes in his hair, I cannot hope to perfectly replicate that, but in pixel art I can approximate it. It will never be perfect, Wittgenstain is made of smooth edges and not squares, as is the tradition of pixel art I'm only using two colors to highlight the pixelness of it.  
2. The flatland. (I actually meant to have more photos showing my progress before this step --will do next time-- but I got distracted. )

This second aspect to art I want to use these early saves to illustrate. Make things flat. People often think they aren't arty enough to get the angles on a room just right and make 'depth' or just can't get the proportions right to make something look... not deformed. But the fact that your eye is able to pick up that you're work isn't 'right' just yet means you have the capability to notice what is off, with is all it takes with diligence and slight alterations to make something looks right. The easiest way to approach this is to flatten things. Drawing still life or things laying around the house is really hard because as you look back and forth you wind up moving your head slightly and because of your perspective the angles on the piece change. You end up constantly altering the proportions of the piece. With practice you can learn to avoid this. But for beginners or those trying to get a bottom up refresher of their abilities(which every artist should do occasionally, imho), start drawing things from photographs. Next is to break the image down into less complex shapes and distances. 'The eye is ovalish, half way down the head, a little squared over here..' ect. This is the same for the negative space. Instead of seeing where to put that tiny ear in relation to all this white, imagine figuring how to fit this large triangle thing between the shoulder and the ear. 'Okay it comes up to a little higher then half the image, bends right below the ear here and slightly closer towards the neck down here.' Start by building your basic outline and move in. Remember to strictly avoid detail work at this point, you're getting the layout of the image. Start sharpening the edges of that layout.

3. Details. Treat every detailed spot like you approached the general layout, start with the big stuff. 'this spot is all black' 'this corner is more rounded' 'this has little hairs sticking up' and then move onto more intricate stuff.
 (As a side note somewhere in the above I said there is no direct technique that leads to how a piece turns out. I kinda lied, the grey stuff is dithering, it's a technique of spacing squares in a checkered or 'cross' pattern to replicate shading. It's a way to keep the pallet color down while simulating shading, similar to crosshatching as seen in comics/print. )

The rest is the back and forth, glance at the reference picture, make a decision how to shade/represent what is there, look at the image see what could be wrong. When you are at this point getting a pair of outside eyes can be very beneficial, also looking at your piece from different angles i.e. turning the image and canvas upside down. Replicating something upside down is a common experiment in early drawing classes and I recommend it, by having a face upside down the brain doesn't so easily recognize it as a face and allows you to more easily see it for it's angles and edges. About here a talented artist friend of mine mentioned how not only was the chin far too thin, but the neckline doesn't go in so drastically, that the whole left shoulder (in relation to viewer) was in too far. Within the next three panels I work on correcting that along with other slight alterations with the eyes and brow. Then carry on to darken the dithering on the other side of the face, make slight alterations to the hair, and work more on the eyes. Click on the first image to be able to cycle through the images without seeing the text to see the changes better side by side.

A great way to break down the shading in a piece, especially a monochromatic, is to squint at it. Or in my case take off your glasses. The details get lost but suddenly the overall shape and shading becomes more clear (in a manner of speaking). A pixel art piece of an image is in a way already a blurred/distorted version of your image, so when you squint and compare the too your eye will better see the features that translate and more importantly where they don't. Try it out: comparing the bottom two, the original image and my piece you'll notice the bottom Wittgenstein's nose is much darker, that there's less shading on his left ear, and that his chin doesn't extend quite so long. With open eyes of course his left eye needs work, it's not clear where his eyeball and eyelid start and stop as it is with the right eye.

You'll notice the comparison is far from 1-1 but it is an approximation. There are light spots I couldn't figure out how to represent near the right side of his temple that I excluded altogether. The dithering more lends itself to suggesting the details and shapes of the face then replicate them. Now the point of this exercise specifically, to attempt to approximate another image (instead of simply creating an image from imagination) is that this teaches the eye how to see in terms of simple things like lines instead of difficult things like faces, and it teaches the mind how to use your medium in approximating certain effects. When your mind and eyes think in terms of the medium you use and you will be better able to clearly imagine a creature in a way to translate it to the paper... or screen.
Until next time, thanks for your time, good luck on your art and goodnight everyone. 

Monday, October 6, 2014

Epoche Development Talk 2, Procedurally Generated and Narrative

Epoche Talk 2, Procedurally Generated and Narrative 

First read, 
Game Focalization and Narrative

First I'm going to give a quick update of my progress. The pixel art program I'll be using for this project is freeware, which can be found here I'm reading into verbcoin now because I need to have a full grasp of how to code my algorithm. With this art program I'm going to sketch up some placeholder art and start the framework of the game. Which means the simplest version of the algorithm. Breaking a many scenario system down to the simplest wheels in a way that it will be instantly playable, but code in a way where the next iterations will fall into place easily (or as easily as possible), is proving to more difficult than said. Particularly with the dialogue section.

(image and a good read found here)

So, what do we always do when we hit a rut? We go around, but we keep digging. Everything in a generated game is connected, but it all stems from a simple set of mechanics. So those mechanics are what I will be working on. The gameplay in each match will be broken down into five stages.

1st, the characters, the game will have to randomly generate characters fleshed out with distinguishing features, stats, and personalized inventories. Setting up a test room and test figures will be my first, and most important, challenge. The quality of the game's mysteries (as currently designed) are dependent on this stage of generation working.

2nd  The basic AI movement of characters through the train cars will be directly linked to their stats. I don't have the luxury of making their movements a scripted pattern as the murders both need to have clue states and as must occur out of opportunity. When hungry the characters will go to the bar cart for a sandwich, which will be added to their inventory, which could leave crumbs on their hands, or the hands of a thief going through their pockets. The stats are a basic excuse for variables within the timeframe of each match, this effectively diversifying the instances of each match. Also by making npc's movements less predictable in later builds the player becomes more reliable on the testimony of the passengers(passive targeting and dialogue farther down). This adds suspense and strategy as the player must diplomat with --potentially-- the killer and put themselves in greater risk.

3rd, the first murder state. The combination of the stat effected movement simulation and targeting's system this is the second make or break point. Making the characters' passive/active targeting will be very very difficult.. for me. As the title implies the targeting code is how the murder selects their victim, this section will work off an expanded build of the first mechanics, expanding the stat and other variables so the game simulates the nature of the murder and the state of the crime scene. This section will eventually decide things like opportunity, weapons available, if there was a struggle, the number of clues left behind, and most importantly the imprint those instances left on the other characters. Which brings us to...

4h, claim dialogue, or better named, claim and counter claim instances. Here is where things get tricky and where my abilities with verbcoin make or break the game, each of the characters will have to have a storage system, exposure to fact states(clues) will allot them with a build of claims. Which with their passive/active targeting will organize how the AI responds to player input. This is the real meat and potatoes of the game. As the player picks up clues dialogue options are opened and they are able to 'build a case' against the npcs to decide which one they think is the murder. But dependent on the clues the npcs will be exposed to a different set of possible 'cases' and possibly against different characters, even yourself. I'm going to have to figure it so that the AI knows to respond, shifting blame(active targeting) to the claim that npc has unlocked and discrediting the claims of those it hasn't come to trust (passive targeting). This is going to be all very complicated, and in this early stage very imbalanced, broken. Which is why I'm going to need more effecting how the npc's make their decisions, how they notice features and build their clue/claim database.

5th, narrative dialogue or essentially randomized characterization/narration, is what I'm calling the extra dialogue I'll need to create. Where claim dialogue is tied to the active targeting, leading to the claims and win or fail states: the narrative dialogue is tied to the passive targeting and will result in the majority of the game's characterization. The passive targeting is essentially how a npc feels about another npc or the player, like a relationship meter, the modification here is that it is modified by the rest of the passengers. What happens is that because of the social situation the npc will assign each character a value from least to greatest (liked) and, dependent on the frequency of positive to negative responses from each character to the npc, assign biases towards those characters. This both effects how the npc will respond to the player's requests but if the npc will 'believe' the player. Certain npcs with a higher rating of the player will need less proof to be convinced of the player's claim dialogue, if that claim dialogue is proven wrong that rated will go down though. The importance of the narrative dialogue is... well the narrative. This passive targeting as an underlying feature that effects the active targeting simulates most of the mystery elements in the game. Not only through what amounts to diplomacy and cunning wit in questioning and organizing the group, but on the side of the murder it'll simulate motive and manipulation. The murders active and passive targeting act the same but as to manipulate the groups understanding of the crimes away from himself, and when the opportunity arises it simulates a murder of opportunity. This, of course, is in much later builds. And currently beyond me, but if the character creations work, and the AI movements work then half the work is done.

Tha'rrs the game plan. The five stages as currently designed, each building on previous iterations, slowly building to the more complete project. I don't have a timeline for the entire thing, I have no way of knowing how quickly I'll learn and be able to implement the code into each build, much less work out the bugs, with all of school and work. But it's definitely possible so I'm going to do it

Next time more news on code and art, maybe even some screenshots. Till then, have a good day everyone, and good luck on your projects.