City Rach: Deepart.io & Exercises

Here’s the photo I want to paint.

I messed with the photo of Rachel in the window in Photoshop a little bit until I came up with this modification of the composition that I think I like best. I won’t print this and paint it, I just wanted to get a quick look at how reordering, stretching, etc might improve the photo’s composition.

The next step is to paint some value studies and start messing with some exercises to practice some of the techniques I’ll need to use to paint this picture. But first, I wondered what it might look like if Joseph Zbukvic painted this. So, I grabbed his painting “Red Awning, Paris” that I found online here. And I used that as the “Style” photo on Deepart.io.

Deepart.io is a pretty amazing little website that takes a photo, and stylizes it based on a Style photo. I have no idea how involved the algorithm is that does this, but the results are stunning, and I’m almost always impressed by how much the result looks like it was painted by the creator of the style photo. Because Joseph Zbukvic is my favorite painter right now, I figured I could use Deepart.io to approximate what he might do with this photo. I thought this might also help me see to the photo in a different way. Instead of just painting what my instincts tell me to do, I can have a computer approximate what another artist might do. This feels a bit like cheating. But, I don’t believe in cheating at Art. As long as I’m not straight up stealing another artist’s work and trying to pass it off as my own, I don’t think there are any rules. So, using Deepart.io in this way is just using the tools available to me to help me see my subject differently. If I went and just painted exactly what Deepart.io generated, then I think that would be cheating. But using it to approximate how another artist might tackle the problems present in a particular composition is not cheating. Especially because, it’s certainly wrong. There is no way Joseph Zbukvic would sit down to paint this photo, and come away with something like the image generated by Deepart.io.

Anway, if you run the photo of Rachel, and Joseph Zbukvic’s painting Red Awning, Paris, through Deepart.io, you get this:

Not exactly as helpful as I thought it might be. I think this is so far from what I was thinking because the style photo and the source photo are of such different subjects. So, I decided to try a different painting as my style source. This time, I used Kitchen Gossip by Joseph Zbukvic instead. Here’s what Deepart.io gave me:

Again, not at all what I think Zbukvic would do with this painting. I guess my idea of using Deepart.io isn’t quite as helpful as I thought it would be. Which, in a way, is a relief – because it means I’m not going to be all that heavily influenced by what a computer thinks a master artist would do with this photo.

One thing I did realize by doing this is that I don’t need to paint much in the windows. In fact – I might end up just leaving the windows white paper, let the viewer decide what is out there… that’s an interesting idea, I’ll have to think about that some more.

Anyway, I played around with trying to paint Rachel’s head. I know this is the single most important part of the painting, so I knew I’d need to practice it a lot. Below are the exercises. Basically, I started by painting light clouds – experimenting with different techniques, and then I painted her head and shoulders. At first I tried to paint solid shapes and thought of shading in the bits and bobs – but as I kept going I realized that it needed to be one color. Maybe I’ll be able to dab in a hint of red at the end, but the head, shoulders, pants, and curtain are all part of the same shape. So instead of painting a head shape, then a hood shape, then a shoulder shape, I should paint it all as a single shape. This means painting the head in one or two brush strokes – and then not coming back ever again. The shoulders and hood and back should all be one tone, and paint around the highlights to use the negative space. It’s hard to explain, just look at the evolution of the head below.

This is also a good demonstration of why these sketches are so important. I need to identify the problems, then try to solve them, then practice the solution. Only then can I paint them with confidence.

No joke – I was trying.
After eleven attempts, I stumbled upon the negative painted highlights – which is how I am going to tackle it in the end.
Then I practiced that a lot more.

All 24 attempts took less than an hour. Each paper is 1/16th of a sheet, and that’s cut into 8 sections. So each sketch is about 2 inches by 3 inches. I also learned that I like a very simple abstract sky the best.

Once I was done there, I painted a value study on another 1/16th sheet. Right now, I actually like this value study.

I’m going to try again before I go on to the full painting. There are a few things I need to work on: I need to paint the head in no more than two strokes. Then I can’t touch it again.

I want to mask the window to get very crisp sharp lines. Everything else will blend and mix together, but the window itself will be crisp white lines.

I need to play with warm an cool tones. And the wall under the window and above the window can lighten up a bit.

I actually like this – though I wish I hadn’t messed her hair up so much.

Be First to Comment

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.