How designs, retouchers and finish artists can embrace AI in their work with Chris Hall

Chris Hall is an experienced designer and retoucher. He's moved from print design to digital and now uses AI regularly as part of his work. In this podcast Chris talks us through:

• How and why he first started using AI in his retouching work.
• How he uses AI to enhance his work (a clue: it's not one and done when it comes to prompting).
• Why his experience in design and retouching help him produce higher quality final assets.
• The limitations of AI when rolling out a campaign across multiple formats.
• Why he had to buy tubes of tomato paste to fix an issue AI couldn't.

Chris also talked through some of his work with Gary and Christina:

• The Allianz eagle, where Chris retouched the AI-generated images so they were suitable for Sydney's largest out-of-home advertisement: https://www.instagram.com/p/DSJQCthkkUK/
• Mutti's Muttiholics campaign, taking the original AI concept so it was suitable for large format advertising: https://www.instagram.com/p/DRQmnIVklYc/ 

Chris' insight shows that even when a concept is strong, a human hand is often needed to bring the AI vision to life.

• Soc•AI•l Skylines is a Social Hills production.
• Hosts: Christina Vetta and Gary Andrews.
• Producer: Peter Northcote.
• Videographer: Imogen Mabin.
• Guest: Chris Hall.

Transcript

SPEAKER_00 0:07

Hi, and welcome to Social Skylines, a production by Social Hills. I'm Christina Bella.

SPEAKER_01 0:12

And I'm Gary Andrews. Each episode we'll be cutting through the noise around AI, speaking to creatives and marketers who are using the tools day in, day out. This episode is for anybody who touches design, whether that's marketing, retouching, or the creative side of things. And I'm really delighted that we're joined by experienced graphic designer and retoucher Chris Hall. Chris, welcome to Social Skylines.

SPEAKER_02 0:35

Thank you for having me.

SPEAKER_01 0:36

I think you're probably one of the first people that I actually saw in the design space using AI for retouching. You're talking about it a lot for design. What prompted you to really start getting heavily into that side of things?

SPEAKER_02 0:50

I've been doing design for a long time, and when I started design, computers became the big thing, which is quite a while ago. And it actually took a lot of people uh out of their work from the old days of paste up to computers. So I realized from an early day that things changed very rapidly. Then there was digital media came in and when I was doing prints, I'd learned digital media after that. And then after that, it was 3D animation and motion graphics. So my whole thing was to keep ahead of the curve. So I could see that AI was the next thing to learn.

SPEAKER_00 1:18

When we're talking about learnings, what has been the biggest challenge with regards to upskilling?

SPEAKER_02 1:25

Normally when you like learn design or um retouching, it's an easy program to learn. It's usually like a step-by-step type thing. But AI is quite rogue, it's almost like it's organic. I think of it like a beast that I've got on a leash that I'm just dragging to where I want it to go. And it could be different each week or even each month because they have updates quite often. So one month you'd be working on one campaign and it it goes the way you want, and then the next month they want the same type of look, but the program might have had an update and it might change quite a lot. That's been quite a bit of a challenge is trying to get it things to be consistent or to know how it works.

SPEAKER_00 1:59

Yeah, absolutely. I mean, when it comes to generative AI, it's very hard to get the exact output. That's one of the things that I've had a challenge with is you know, you kind of get on a track, it just falls apart because it doesn't remember exactly where it was.

SPEAKER_02 2:14

It's consistency and it's just a lot of the times that you know a client will start talking to you about how you're going to do this, and you start thinking, okay, this will be an AI job, and then you start to panic because you don't know what's going to happen ahead with the AI because AI can go on a tangent, like you said, and you can waste hours if you're not in the right way. So it's like a beast to me. It's just like pulling this beast around and making it get to where I want it to get to the finished product.

SPEAKER_01 2:35

Talk us through a couple of pieces of work that you've done. How did you use AI and why did you decide and where did you decide to use the AI?

SPEAKER_02 2:43

So I recently finished a job for Mooty Italian sources. The client came to me, it was actually the agency came to me where they needed help. So a lot of agencies are using AI to come up with concepts, but those concepts aren't production ready. So they need someone like me to actually take it to the next level to elevate the work and to make it production ready. This Moody campaign was hard. There were three characters in it, three different moody characters that needed to be done. The visuals are really nice from what they sent me, but just need to be more realistic. So it's tricky to get it to that point. Plus, I had to work in a large format campaign. So did all the lighting. There were three characters that had to be consistent with lighting and how look and feel, so they worked across the campaign. But uh when it came to actually like the products and involved in this actual pitches, it didn't quite work very well. There's one in particular, there's the an older female, she's got hair rollers in her hair, but the instead of hair rollers, it's the product. The product needed to be on different angles. So when I was generating the AI to get these products on this different angle, you just keep doing the same angle. So that's kind of a thing with AI. It predicts what it the image will look like. And a really good example is if you ever do a test with an analog clock, it always does the hands at the same place, so it's always 10 and 2. And 10 and 2 is because the AI thinks that all analogues clocks look like, but it's because they've been trained and scraped the internet to see all blocks are photographed on the angle. Whenever a photographer takes a photo of a clock, they do it on the angle, it's just like the thing. So if you type in a generate block, it will do that. With the product, keep putting the product on the same angle, which was like a front three-quarter angle, because it's seen that photo so many times, it thought that was the correct way to put it, just like putting a tail on a tiger.

SPEAKER_00 4:23

It's pulling in the what it's learnt.

SPEAKER_02 4:25

Yeah, so I realized very quick. I tried to do sketches of it on different angles of the actual product. I tried to prompt it by saying go lower, go on this angle, go getting very annoyed by now. I was like trying to get it to go the way I was pulling that beast, but then I realized quickly I had to go to Woolworths and actually take photos, get the product, buy the product. Thankfully it was already out and I took photos of it, and then I comped it into the shot with my retouching experience and got them all in a different angle. That's where it works. So that's the limitation of that. Wow.

SPEAKER_00 4:54

Just on that, how do you manage the client's expectations around the use of AI? Because I think there's a misconception that agencies, designers, creatives, when they're using AI, it's making their job easier. Yeah. So you've just kind of spoken about a whole bunch of processes that you've had to go through to get the best possible output. How do you prompt the client or whoever you're doing the work for that this is actually going to take a lot of time and there are a lot of unknowns around this process that I'm just going to be up front with? How would you manage that?

SPEAKER_02 5:32

Yeah, this is a tricky one. Everyone's got the capabilities of using AI. You can use it on your phone. You can generate these amazing pictures, but they're never production ready. So that that's the limitation is telling the client there's more work that needs to be done. A really good example with AI going rogue and wrong is and I don't know if you've seen it in London, they've done a massive mural. Yeah. Christmas mural. No, I haven't seen that. Okay, so this mural is huge. It's in this West London kind of above a restaurant. And they've obviously just pressed the button and AI generated a like a scene of uh some kind of like festival in the medieval times in London on ice. They've just gone and sent it to the printers, and the printers have blown it up and put it on the sign. And then everyone has realized that in the image there's dogs that have eagle heads, there's body parts coming out of snowmen, there's faces coming out of other people's faces. And if you know anything about AI like I do, you have to actually you zoom in and you can see all these really obscure things. Some are quite scary, some are quite random, and so you have to know that can happen. Clients do think that it's just a click of the button and it's ready to go. It's not, it's just a start. There's a whole process to get to where you want it to be production ready in the real world environment and work in a large format type of environment. So it's just letting them know that hey, this is great, but that's my starting block. I need to do a lot more work. And a lot of it's old school methods of retouching rather than just AI generation.

SPEAKER_01 6:53

You've talked about a couple of limitations there, which is the clocks that don't actually show any other time. You've got eagle heads on dogs which will give me nightmares. Are there any other kind of really major limitations that you found when you're using AI?

SPEAKER_02 7:09

The biggest limitations, kind of what I touched on before, is how large these generations are coming out. When you generate an image, it's quite small. Sometimes they can be used okay for a social media campaign, but when it comes to actually large formats, a lot of the stuff I would normally do would I'll be given a whole range of photographs taken from a photo shoot. When it's an AI created thing, it would be more smaller image that needs to be uh taken to the next level. A really good example would be um I just worked on uh campaign for Alliance Insurance. I don't know if you've seen the Eagle Ab with the third on the back.

SPEAKER_00 7:43

You worked on that?

SPEAKER_02 7:44

Yeah, the uh the actual still static. That is such a cool campaign.

SPEAKER_00 7:49

Love that.

SPEAKER_02 7:50

They had the TV C, which just looks amazing, everyone loves it. But the problem was they didn't think too hard about how they were going to produce that in a static environment and outer home, like on billboards. It turned out that the actual Eagle render was not at that next level. It was quite chunky, and if you look at the ad, you can kind of see it. It wasn't finessed, it didn't have texture, it didn't have real Disney features. They came to me with the problem of this 3D image: can you make it look more realistic? And that's something I hadn't done before. So it took me a while to get my head around how to do it, but I ended up doing it in AI. That was the first challenge. The second challenge was making it actually work in a large format situation. So the second time they came to me after the first campaign, they needed a one-off slot for the largest bellboard in Australia, so the Glebe Silo building. They wanted the eagle to be on a different angle and a really nice moment where the eagle's looking at the bird on its back. They wanted that on this bellboard. Now, when you're splitting out an AI image that's like the size of an A5 and you go into something that's 20 meters wide, you need to think how is this going to work? So I've got a method, a way of actually making that work with my experience over the years of print of how to make that large format. That's a real limitation at the moment with these AI-generated images from clients. They're not big enough to use in an out-of-home campaign.

SPEAKER_00 9:08

You sound like you you've nailed it though. And that's the whole thing with AI, is everyone's figuring this out in real time.

SPEAKER_02 9:14

Yeah. The way I do it, the method I do it, it's not just one image, it's a lot of different images stitched together. But it's also my re-catching skills that's come through. I'm using AI like a stock library, but a really sophisticated stock library.

SPEAKER_01 9:27

No, that makes sense. And to be honest, like I've seen the Mutti campaign out and about in some really big billboards, and until you told me it was AI, I had absolutely no idea whatsoever into there. It looks really, really good. One of the final questions that I want to ask you is everybody thinks they're a designer, everybody thinks they can press a button and know where to go. But for a lot of people who are looking to upskill and perhaps, you know, got basic design skills and want to kind of push it a bit further. Are there any apps or any tools, anything that you'd say this is a really good place to get started?

SPEAKER_02 10:02

With the world we live in now with social media, I follow a lot of influencers or other people using the same technology and I follow them on Instagram. Um, and then if I'm happy with what they're doing, I'll follow them on YouTube and they do a lot of tutorials. So it's just taking a bit from here and a bit from here and just trying to work out which is the best way to do this kind of stuff. But YouTube's amazing for learning. They will go through how they got to a certain thing. That's pretty much how I've done it, and just a lot of trial and error. But yeah, also with the years of working, I've just gained all this knowledge of retouching that it's kind of like I forget that I know all that knowledge, but I that's where I would go. I would go to YouTube.

SPEAKER_00 10:39

Thank you so much for joining us today. We've loved your guidance. The work that you've produced is fantastic. The Moody campaign is awesome. I actually use the product myself. So when Gary showed it to me, I was like, this is awesome.

SPEAKER_01 10:52

You don't have the rollers in your hair though.

SPEAKER_00 10:56

If anyone wanted to get in touch with you, how can they do that?

SPEAKER_02 11:00

So my email is Chris Hall at gustoevents.co. They could look me up on Instagram and they'll find my page there as well, or LinkedIn, Chris Hall, graphic designer, and they'll find me.

SPEAKER_00 11:11

Awesome. Well, I'll definitely be picking your brain after this. Yeah, yeah.

SPEAKER_02 11:14

I've got a few more to drop, like on um social media. Yeah. I just haven't had time to do it because he's so busy.

SPEAKER_01 11:20

Awesome. Well, that's fantastic. Well, thanks again, Chris. Really, really appreciated some fantastic guidance there. I've learnt a lot as well. Don't do clocks using AI. That's probably my big takeout. Thanks so much, Chris. And you can find social skylines on your podcast app, whichever one you use. And you can also follow us on social channels. There's the special course website, there's Instagram, there's LinkedIn. So please do follow, please do like, please do subscribe. And we hope you find this useful.