RatCreature (
ratcreature) wrote2008-05-26 08:15 am
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
Entry tags:
fanart rambling...
When I was drawing my latest piece, i.e. the SGA/Avatar fusion with Teyla as Waterbender, I was reminded again why I'm rather reluctant to try drawing fanart for tv/movie fandoms, my recent forays into SGA notwithstanding: I have a hard time with character-likeness if the character has to look like a real person.
Because my style of drawing is more comic/illustration-like than truly realist, e.g. that I like to have lineart, it needs a certain amount of simplification in facial features. Which then presents the problem of how to get there from the starting point of a realistic and fully rendered face. (Not that I can do realistic portraits, but in theory I mean.)
The first thing that usually comes to mind for trying to get a handle on how a character looks is to start with a photo of the character's actor or a screenshot of the character, and then somehow simplify from there. The reasoning is that after all basing your art on a decent photo works well enough for realistic character portraits in fanart, which are often recognizably based on promo pics and such. Yet this approach is somewhat hazardous as anyone who has seen a bad tv comic, one where the artist visibly just traced screenshots, can attest to. It's the phenomenon that in its extreme is lineart that you could even actually map over a screenshot and the lines "fit," yet if you look at the lineart alone it doesn't really look like the character at all.
The problem is of course in the nature of lineart. If you have ever tried to trace a photo, you've run into the problem that there aren't really any "lines", so usually you tend pick mostly the "high contrast borders" with a bit of abstract knowledge of how the form of the thing is thrown in. And this works okay if you have say the contrast of a leg against a bright background, but much less for things like facial features. And it is not merely distortions due to a specific photo, i.e. that depending on the light and angle your best guess for lines may not emphasize the really prominent features, but put stress on the wrong parts. It's that any reduction of photos to lines with a face makes it a caricature, even if you don't add intentional "distortions," simply because having just one line where there used to be color gradients introduces emphasis, and likeness decreases if you put that emphasis "wrong", i.e. not on the recognizable, outstanding features.
In theory this is not much of a problem, after all the goal all along is to draw the character, not to trace photos, and you just have to adjust your degree of caricature to compensate for the reduction of rendering, that is to figure out which facial features of said person deviate from the average proportion, the mean of facial features in a way, and exaggerate. I've read that even computers can do this with algorithms based on photos and make caricatures of people.
The problem I'm having is that so many actors are pretty people. See, I'm not that good with faces. It's one thing to spot how someone differs from "average" if they have huge ears (think all the Prince Charles caricatures), or a big nose, or a very distinct skull shape, but humans tend to find regular, even features more attractive, so tv characters are hard to figure out. It's not that there are no differences, obviously I recognize these people when I see them (well for the most part anyway, like I said, I'm not that good at memorizing faces), but I have no idea which features are the ones standing out most to me on a conscious level with faces like that.
I think it would be really cool if one of those caricature algorithms was made into a webtoy somewhere, and I could just give it a photo and it would warp the features to point out how it differs from the average face. Then, even though my style doesn't need outright caricature, I could use those hints for more subtle exaggeration suited for my purposes.
I guess I just wish some technology could help make up for my lack of talent/practice in character portrayal/caricature. *sigh*
Because my style of drawing is more comic/illustration-like than truly realist, e.g. that I like to have lineart, it needs a certain amount of simplification in facial features. Which then presents the problem of how to get there from the starting point of a realistic and fully rendered face. (Not that I can do realistic portraits, but in theory I mean.)
The first thing that usually comes to mind for trying to get a handle on how a character looks is to start with a photo of the character's actor or a screenshot of the character, and then somehow simplify from there. The reasoning is that after all basing your art on a decent photo works well enough for realistic character portraits in fanart, which are often recognizably based on promo pics and such. Yet this approach is somewhat hazardous as anyone who has seen a bad tv comic, one where the artist visibly just traced screenshots, can attest to. It's the phenomenon that in its extreme is lineart that you could even actually map over a screenshot and the lines "fit," yet if you look at the lineart alone it doesn't really look like the character at all.
The problem is of course in the nature of lineart. If you have ever tried to trace a photo, you've run into the problem that there aren't really any "lines", so usually you tend pick mostly the "high contrast borders" with a bit of abstract knowledge of how the form of the thing is thrown in. And this works okay if you have say the contrast of a leg against a bright background, but much less for things like facial features. And it is not merely distortions due to a specific photo, i.e. that depending on the light and angle your best guess for lines may not emphasize the really prominent features, but put stress on the wrong parts. It's that any reduction of photos to lines with a face makes it a caricature, even if you don't add intentional "distortions," simply because having just one line where there used to be color gradients introduces emphasis, and likeness decreases if you put that emphasis "wrong", i.e. not on the recognizable, outstanding features.
In theory this is not much of a problem, after all the goal all along is to draw the character, not to trace photos, and you just have to adjust your degree of caricature to compensate for the reduction of rendering, that is to figure out which facial features of said person deviate from the average proportion, the mean of facial features in a way, and exaggerate. I've read that even computers can do this with algorithms based on photos and make caricatures of people.
The problem I'm having is that so many actors are pretty people. See, I'm not that good with faces. It's one thing to spot how someone differs from "average" if they have huge ears (think all the Prince Charles caricatures), or a big nose, or a very distinct skull shape, but humans tend to find regular, even features more attractive, so tv characters are hard to figure out. It's not that there are no differences, obviously I recognize these people when I see them (well for the most part anyway, like I said, I'm not that good at memorizing faces), but I have no idea which features are the ones standing out most to me on a conscious level with faces like that.
I think it would be really cool if one of those caricature algorithms was made into a webtoy somewhere, and I could just give it a photo and it would warp the features to point out how it differs from the average face. Then, even though my style doesn't need outright caricature, I could use those hints for more subtle exaggeration suited for my purposes.
I guess I just wish some technology could help make up for my lack of talent/practice in character portrayal/caricature. *sigh*
no subject
I don't know how helpful this is, but I read something once about how sometimes it's useful to practice drawing caricatures by drawing people who already have sort of "exaggerated" features? Like Angelina Jolie or Arnold Schwarzenegger? Plus then, I guess, it would be easier to find other caricatures to compare your own efforts to.
no subject
no subject
I'm not saying it's a fantastic way to draw in general, but for getting TV character likenesses, it's actually very effective to draw a general "pretty" face and surround it with the character's usual trappings. As long as nothing is blatantly WRONG -- for example, John's face is kind of pointy, and a square lantern jaw would look out of place, but pretty much any narrow anime-style face looks more or less like him.
Edit: Which, come to think of it, is actually a feature of caricatures in general. You don't caricature everything on their face and body -- you do a generic face/body with exaggerations of the two or three features that are most distinctive, whether it's their hands, their nose or their hair. A totally generic face with very distinctive hair is still a caricature if there's nothing striking about the original person's face.
no subject
no subject
"I've read that even computers can do this with algorithms based on photos and make caricatures of people." Really?! How cool!
no subject
But yeah, I've read about the caricature programs on Wikipedia:
http://en.wikipedia.org/wiki/Caricature#Computerized_caricature_and_formal_definition_of_caricature