Lynch dancing girl

[Zizka]

Hello guys,

I'd like to do an animation of the girl in Mulholland drive dancing.

ref:

hebergeur dimage

V1:


V2:


I'd essentially to get a third version which will be more readable so that I can start animating it.

If you provide criticism I will do my best to put it into practice.

Thank you for reading,


Z.

[astraldata]

Decided to give this one a shot:

and

Edit includes the following:

-12 colors
-higher contrast
-more blended features
-more shapely forms
-now works with any BG of any color contrast
-truer to reference image's forms and highlights


---


Granted, at this resolution, an image like this is more of an approximation than a copy, but the main flaw in your previous process was that you were symbolizing your forms rather than approximating them as best as possible. This involves taking into account the actual form and lighting, according to your reference, as best as possible. A good example of this issue is when you converted the eyes and eyebrows to simple dots and lines in your first image. Your second one was better, but this still didn't take into account lighting intensity and proper contrast very well in your form representations. The bridge of her nose, for example, just reads as completely flat in your second image.

The best way to solve this is type of issue, of course, to have proper contrast (but NOT necessarily more colors) in your palette so that you can better represent detailed forms while taking into account the lighting. You may want to work at a lower color-depth in detailed images like this one so that you can globally tweak the intensity of your colors at an instant whenever you need to in order to see whether you need more or less contrast with a particular color.


---


Another issue is that the shape of the face/head does not closely resemble the source image shape. Her face is more round, though your representation is more of a longer face. In your second image, think of Matt Smith.

The best way to think about representing lighting on details like cheekbones and the forehead is to think about lighting them simultaneously across the face in a pool of light. The places where there are dips may or may not deserve a shadow color (irregardless of the reference image stating that there is indeed a shadow there) depending on how important the detail is for clarity. This is where you get into approximation in your light and shadows.

That said, think of shadows as shallow pools of water (where lighter things can emerge from) that just blob together in the darkest areas of the form.

The way to determine whether you need a shadow pixel in a place has to do with where the rest of your shadows are pooled and how many levels of graduation you have in your palette. If you have the necessary palette to indicate a darker shadow in places that need it, go for it -- otherwise, pool your shadows together wherever important details are not required to emerge from the shadow, and remember that each level of emergence from the shadow (each shade in your ramp essentially) is a different depth or layer of shadow. In the case of an important detail, allow it to emerge from the shadow 'pools' only just as much as necessary, leaving the rest of the shadow in place, and possibly buffered by other levels of shadow to smooth the transition back to darkness if you want a more rounded form (think of it like 3D height maps I suppose).

You're essentially pulling the form out of the darkness with light, so contrast becomes your best friend.

[Zizka]

Thanks for the feedback, astraldata.

The bridge of her nose, for example, just reads as completely flat in your second image.

I've reworked this.

As for the rest, I've used your edit more than the text. I foud the text interesting to read but wasn't too sure as to how to apply it.

Astraldata:
V1: V2: v3:

This is the link of the video in which she dances and which I'm trying to reproduce:
https://www.youtube.com/watch?v=VHPKe8D01Kk


Whhheeeeeeeee!

[Decroded]

i started having a go but ran out of time 

[Zizka]

@Decroded :

Thank you very much for the edit.

After analyzing it:
-I need a rounder head (this was mentioned by astraldata but I only realized it with this edit for some reason ).
-Most importantly: darker hair color. I double-checked my ref after looking at the edit and the hair in my version is indeed to light.

I have a few comments/questions about the edit  .

*I noticed you put eyebrows there while I didn’t. The reason I didn’t is because, considering the scale, I didn’t think the eyebrows were “worth” a whole single pixel. I’ll give it a shot and see how it looks.

*You changed the light source, am I right? I don’t understand how the neck would be lit that way, I mean on the lower left, where it’s very bright. Wouldn’t the head and the chin cast a shadow on the neck? I’m not saying you’re wrong. This is a question to further my understanding, not to say you’re wrong. I took a pic of the video and this is what I got. The arrow points to what I’m referring to.

 
*I wasn’t under the impression she was tilted a bit like this perspective wise. The color of the neck and on the side of the face really makes it stand out in your edit. Is it because that’s how you saw it or because you wanted to give it a more interesting pose by not putting her straight on face forward? Again, a question to improve. 

*I also notice you use a lot less colors than I do for the face. The face is more readable in your version. I feel like you went for readability as opposed to going for something which was identical to the ref.

*Also, you didn’t put the sparkles everywhere on the dress. This is what made me think that you put the light source on the upper left as opposed to right on top. I take it you put less sparkles because you felt it looked too busy otherwise.

I like your edit, I find there’s a special touch to it . It has a cartoonish twist to the ref in my eyes and at this resolution I think it might be a good idea.

Not related to the edit, there’s something my eyes have trouble seeing in the ref: the volume of the dress.
As you can see in the first two versions, it was very flat. Astraldata pointed me in the right direction with his edit but, although I could see more volume in his version, I couldn’t see what he had seen in the ref picture.

If you look at the following picture (once again taken from the video I linked in my previous post):

 
I have trouble seeing how I can recreate more volume here. I’m talking, for instance, about the transition from her breast to her belly.

[Decroded]

Umm for most of the questions u ask, the answer is because I stopped suddenly and had to go to work.
Didn't even get to draw the collar let alone finish dress etc.
As far as the style and choice to put eye brows I dunno, not conscious decisions.
I never rly do portraits (which I why I did this for practice) but if I did I would go for maybe about double this scale or at least 50% more if that was an option.
I think that would be far easier and be about the base res to really show her character.
But that's probably personal choice based based on not having such great skill at low res 

[Zizka]

Ahh ok! I see. Thanks for the reply.

So I used decroded’s edit to improve my own in a few things: number of color for the face, color of the hair but most importantly: the dress. I think it looks much better now. Also, if I compare my first attempt with my fourth one, I’m happy with the progress.
I think it’s good enough at the moment to move on to animation unless someone else would like to add something major which they think I should tweak, I’d go with that version.

[PixelPiledriver]

Yo dude.
Long time.
Where you been?

Watch your contrast / key.

[Zizka]

Hello my pixel sensei.  Digging the new avatar.

Woozah, having a hard time with the contrast here.

V.5:
-Made the shoulders rounder so that she doesn’t look like a football player.
-Redefined the shapes of the arms.
-Reshaped the collar to better represent the reference.
-Made the neck slightly shorter.
-Redefined the hair a little.
-Made the nose shorter.
-Accentuated the contrast further.
-Added the microphone
There’s still some readability issue but I think that overall it’s improving.

[dpixel]

Tried a different approach with an outline.  Unless it was on a dark background, it was getting washed out to me.
In my edit, I made her 1 pixel wider to get slightly more detail on the face.  The one dark pixel eyeball wasn't working for me.
Also the pinks from the dress seemed to be a good complement to shading on the face. 
12 colors: