The recent update allowed for brushes which would create buildings and structures – I thought maybe the program could create some structures to put into the landscape.

• 

• 

• 

• 

• 

• 

Very intriguing results – not quite what I was looking for.

I took the generation I used for the previous oil painting test and got the image printed onto canvas – I wanted the output to be landscape format but the image output is square, so I decided to crop in – that way there is a full ‘underpainting’ base created by the computer.

• 

• 

The image is also very low resolution and quite undefined/blury/pixelated, so scaling up to a larger size could be problematic.

The final result was much better than I expected! It doesn’t really look obviously pixelated at all – more just blurry. This should work amazing as a very detailed underpainting – If I just work back in a define areas then it will start looking quite clear.

I was curious as to what would come from putting my final artworks back into the program

• 

• 

• 

• 

Interesting results, but not really more engaging than previous examples – a bit more normalised.

I had a meeting with Tanya and talked about where I am going with my final output.

At this stage I had the concept of created a number of different output variations, done reasonably quickly and then one final big one which would be rendered in greater detail.

Notes from discussion –

• Working in physical (oils) lends more to one final image – digital lends towards variations
• Develop the ‘pipeline’ – Inputs, multiple steps? Name generator, colours (illustrator)
• Colour is lacking in created pieces – could be more interesting
• Feeling quite normal? How could it be a bit more engaging and out there?
• Big painting – show design process for landscape, creatures, plants etc. Final output more fine-art/illustration – Wayne Barlowe.
• PRINT ON CANVAS? – The computer element is feeling a bit lost, and the generated images are quite beautiful – almost standalone artworks. Great textural base.

Moving forward

I definitely want to do a physical piece. We decided that the best route may be to create one big final piece and show off the creation process and steps that went into the collaboration. This would show concepts for characters, plants etc.

Some artist reference she suggested

Justin Sweet (left) & Vance Kovacs (right) both have great textural grounds to their work that would work quite nicely in my piece – the AI generated imagery creates quite a bit of texture

• 

• 

Nick Keller – Looking at conceptual drawings/developments – showing the concept behind pieces that would go into the final artwork.

J M W Turner (Left) , gauGAN generated artwork (right)
The artworks created by the program are sometimes quite similiar to the beautiful textural artworks of classical painter Turner, who created more ‘abstracted’ landscape paintings with amazing depth of texture and colour.

• 

• 

I went back on to the gauGAN program and found that there has been an update since I last used it – The program now has more brushes available for use.

I had a moment of panic as I found that my process no longer worked like before, it didn’t seem to recognise the templates created by my program. After a lot of trial and error I found that I needed to make sure the size was exactly 512×512 (it was 500 x 500) and that there were absolutely no colours that weren’t meant to be there – the anti-aliasing on the lines was creating variation in colour that gauGAN did not recognise.

The program now has different categories of brushes – with the addition of some structural elements which could be quite interesting.

Updated the generation so that the lines are jagged – no variation in colour.

• 

• 

Testing doing a full reasonably detailed painting in oils to see if I would like to work in physical medium for the final output.

I decided to repaint this piece that I did digitaly for supercrit to see the difference.

Starting with acrylic underpainting

Beginning to block in oils

Completed painting

As this was a test I didn’t go to crazy on the details, and I have left out the extra elements I introduced in supercrit such as the animals and extra plants – so the scene looks a bit empty. I kept this one quite small – the final I think would be very large format. Shouldn’t increase the time by an insane amount as I would be painting for it to be viewed from a distance, so not too crazy detailed.

This was however a much more successful test than I thought. This painting was really quick and I am super happy with how it turned out. It probably only took me about 5-6 hours total. I would say the digital test that I did previous took much longer and didn’t look as nice.

I felt so much more confident and comfortable working on this piece, which makes me think that physical is the way to go.

It is getting near to exposure time and as such I have to start considering how I frame and talk about my project – this is what I submitted for the first exposure form.

My project gives a look into the immense potential that can be achieved through collaboration with AI when creating art. I have explored this through the context of concept art, to conceptualise and illustrate a selection of possible fantasy worlds. The artworks are created in a process of co-creation that allows the computer equal agency of the process of creation, and fosters an enhanced level of creativity and imagination.  

Having a go at a full painting of one of the scenes that I generated previously using the process I used for super-crit. To create this painting I used the technique of Matte painting which I researched in my previous post – this is a combination of photo elements/textures and digital painting

The generated image –

Photos used, taken from google and a photo reference pack.

• 

• 

• 

• 

• 

• 

• 

• 

First, I created a greyscale composition mockup/underpainting.

Process –

Final Painting

Reflection

This process was quite new and reasonably challenging for me, I definitely enjoyed learning a new technique and exploring the possibilities there. Im still finding it to be quite messy – I still don’t really know my way around digital painting very well. I found that I relied to much on the photographic reference and let that decide a lot of the composition rather than sticking to what I had decided.

This is quite a relevant way of painting if I am thinking in terms of concept art – Photos are used often to speed up the process. And it does make sense in terms of the final image being a mix of machine made imagery (camera taking a photo) and hand painting. One of the big reasons I am interested in taking a digital approach is that I have always wanted to become more proficient at it and this could be a great opportunity to explore that.

The final result is looking quite interesting and although it still needs a decent bit of work I am happy with the result.

Moving forward I want to try and create a much more refined version of both digital and physcial – First I am looking into some different techniques for painting digitally. I am quite rusty in digital so I looked back at some tutorials by concept artists – Shaddy safadi is a great concept artist and someone that I used for learning digitial painting a while ago.

I am also looking into matte painting techniques – this involves overlaying photo textures and combining with painting to create a photorealistic landscape painting. This is a technique used very often in the industry to create great looking concepts much faster than through hand-painting every detail.

Both tutorials had a strong emphasis on starting with a blocked in base – like an underpainting/sketch, figuring out the composition and some basic values.

This is something that I was lacking in my previous paintings, the computer creates an interesting scene but I need to break free from the lines and create a more dynamic and engaging composition.

For the supercrit advising session I talked to Tanya. She was quite excited about the direction and potential of my project

Notes from feedback –

• Generated name – how would others interpret the name and how might that influence how the outcome would look. For example when Tanya saw the name she expected a nordic type of landscape.
• Human data – how would it feed in?
• Design around the shapes – don’t just stay within the confines. Iterations/developments etc.
• Different process for creating creatures vs landscapes?
• Physical medium could be a nice contrast.

I came out of this feedback session with a pretty good idea of where to head over the break – I need to focus on the technical execution of the artwork and how that is all going to look/ be done.

So moving forward I will create a couple of new worlds – one will be digital and the other physical. I want to spend a good amount of time on each, rendering them nicely. From here I will be able to chose which medium feels right and works the best. Digital fits well within the project, and is most common for concept artworks, but I am much more experienced and comfortable with physical. Physical also could be a great final piece as it is breaking away from the computer and creating it within my world, but still very much decided in collaboration with the computer.