Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

Everything Cambridge

Everything Cambridge artwork

Brian Naughton - Ai Generated Photography

Aug 22, 202336 minAudio
ShareXLinkedInFacebookWhatsAppEmail

Episode 3 - We talk about the hottest topic, Ai generated images and photography.Brian is a writer, artist, photographer and musician from Cambridge, UK. He has been an enthusiastic early adopter of technology since his teenage years; when, as a drummer, he became interested in the arrival of drum machines and samplers. In a similar vein, as a lifelong, passionate photographer (and with an MA in Photography), he recently decided to embrace the new crop of generative AI tools for image creation (rather than seeing them as a threat). Brian is currently working on putting together an AI Photography website to showcase his new work, but in the interim you can follow his AI work on Instagram - @naughton.photo Since then, SNAPSHOT has grown - and it’s about to change. In the coming weeks, we’re taking SNAPSHOT to the next level: expanding our focus from photography to the wider worlds of cinema, visual art, and the people who shape how we see. If you’ve been with us since the start - thank you. If you’re just joining - welcome. What’s coming will be bolder, deeper, and more ambitious than ever. Discover more about our production company and creative work at: https://www.cambridgechronicles.com - Narrated by Domininkas Zalys

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show. Tap a time to play from there.

Welcome to the third photography therapy podcast with Brian. Could you please give us an introduction of who you are and what your background in photography is please? Sure, yeah. I started developing an interest in photography when I was a kid and my dad was a keen amateur photographer and he had a Hasselblad and a Leica M6 which he had in a locked filing

cabinet which I didn't really have access to and I was extremely fascinated by these objects and then he also had lots of photography magazines. I don't remember exactly the names but I do remember sort of pouring through black and white documentary photography and stuff by Henry Cartier-Bresson and the like, Magnum photographers. So yeah, I developed a sort of an early interest in photography and then it's something as

a hobbyist that sort of lasted throughout my sort of teens and early twenties and yeah, just sort of continued as one of the things. I wasn't trying to make a living out of it but I had sort of various successful sort of competition entries and so on and so forth. So yeah, it's just been a sort of a regular fixture in my life really, taking pictures, travelling with a camera.

I wasn't technically trained but I did go and do an MA in photography and Brighton as a kind of treat to myself in 2015. Yeah, so it's just been not necessarily a way of earning money although I have done some professional stuff but yeah, just a constant fixture in my life really as an artistic passion I guess you'd say. Thank you. So you've been involved in photography pretty much all your life as you just said now and I've known you for years, years and years and I love your documentary stuff.

And can you tell me and us the transition between real life photography and AI photography? What were the elements that made you switch? What are the elements that you like and just give us an overview of how does it work and what it is because I'm sure there's thousands of people that they don't even know about the existence of it and if they know about it they don't know how to use it or what it is. Sure. The first thing I'd say is that I haven't really switched as such, I'm still considered

myself to be a photographer and also AI photography which some people are calling post photography is a very new development. There was a couple of breakthrough products last year, one called Dali which was launched by a company called OpenAI which Elon Musk funded and another one called Mid Journey and another one called Stable Diffusion and what these are based on something called

large language models which was a form of AI that has enabled a form of natural language sort of expression so that you can express yourself in natural language and the machine learning and AI components can understand what you're saying. In terms of imagery this relates to something called text to image AI and what that means is that you can describe something that you want to see and the AI will make its best

effort to return output which relates to your expression. So the first I got on a wait list for Dali quite early sort of towards the tail end of 2022 and the first prompt I did was an ice skating polar bear in the style of a Renaissance painting and it came back with this incredible illustration of a polar bear with a kind of like a almost like a tutu like a ballerina's skirt on ice skates in the style of a Renaissance

painting so yeah I was hooked from then but photography at the beginning was something that was just not up to scratch but Dali to some extent has got better at photography but Mid Journey is the real winner in the photography space in my opinion and they've gone through several iterations from the original version version 1 to version 5 which is the present version now version 5.2 and each time it just gets better and better so the returns

if you know what you're doing in terms of the prompts that you use are becoming more and more photorealistic and I've just been kind of inspired to explore this new domain and amazed by the results excuse me and I've been publishing the results on Instagram and I've actually been getting more interest than I was getting from my own personal photography which is interesting. And where can people find your Instagram?

My Instagram handle is nortonnaughton.photo. Just in case you're listening and you are, go and follow Brian and Brian can you give us an example of how AI photography works and how do you generate an image? So with Mid Journey it's all based on a rather bizarre kind of interface which is based on a Discord server so you have to set up an account with Mid Journey I think it's midjourney.com

and then once you've signed up you have to also have a Discord account which is free so you go to their Discord server and then you type slash imagine and it brings up a kind of a prompt interface then and you put your prompt in and the prompt could be very simple it could be a portrait of a cat but as you sort of learn a bit more about the

prompting environment you can become sort of more and more technical for example you can use cinematic language to describe the kind of viewpoint and the kind of shot so if you want a fine art portrait which is kind of like head and shoulders there will be one kind of expression you know various expressions that you'll use but you can use the kind of expressions that filmmakers might use in terms of establishing shots you know whether it's

wide full length half length etc close up extreme close up and you can also use technical terms to generate them and you can also use various cameras like you can I've experimented with Leica cameras quite a lot so you could say a Leica M10 monochrome or you could say a Leica M10R or you could say a Hasselblad X1D or a Sony A7R5 or anything

in between you can also specify lenses and apertures so and you can put words in like bokeh if you want to get you know certain quality to your images it's slightly hit and miss but the idea is that the AI knows what through the training of its data set its image set it knows what a photo made by a Leica looks like and it knows what a photo made

by a medium format camera looks like so it will try and approximate that based on the prompts that you provide it with. And where does the good information from is it just the internet itself or is there like a particular way of knowing how a Leica M3 looks like? Yeah so I mean I'm not an expert in this but they're trained on huge data sets and so that presumably would include lots of metadata for images both images that have been selected

by the companies that make these systems and also images off the web basically and there is actually a bit of a backlash in terms of copyright but there are some cases ongoing at the moment where certain artists have filed cases against some of these AI companies because they've been trained on their image bank without their consent and that represents a kind of

a breach of copyright so case law hasn't been established really but there are some interesting cases working their way through the courts at the moment. Yeah so I kind of visualise it as being a bit similar to the kind of hip hop and music sampling revolution you know there's a lot there's part of this is to do with remix culture and part of it is to do with appropriation of other people's copyrighted work and there needs to be some kind of balance found.

I mean I don't think that if you're describing you know if you're saying an image in the style of Martin Scorsese that that should present a copyright problem but if you're creating a derivative work which is very specifically based on an artist's original conception then you know there has to be some balance struck I think between the existence of intellectual property and the capacity for people to be creative with these new tools but it's a very

new I mean it's less than a year old so all of this stuff is still sort of up in the air. This wasn't actually one of my questions what do you think about? Alrighty let's take a break. The Cambridge Photography Show is happening this October if you're interested in photography, videography or about gear you should come and visit because all the major brands are there loads of photographers attending to network, great variety of talks and workshops you got a lot to miss out so if you're interested just please visit CambridgePhotographyWeek.com

once again. CambridgePhotographyWeek.com to find out more about the show, about the tickets and what's happening. Now back to the podcast. The art belonging to the artist and so on and so forth so this covers the question really well I mean yes I agree that it has to be a certain balance because if I'm an artist like nothing case of photography because I think it's really really hard to claim your rights of an image. Whereas an artist I think there should be a middle ground somewhere whether it's within submitting your images wherever you submit or I don't know making a case to really one

of these companies but as you said it's the beginning so there's that much you can do about it. Yeah I mean I think basically the precedent will be set by court cases you know as it is in so many other areas that once judges start ruling in individual cases that will start to set up a you know like a precedent in terms of case law that other people can refer to so yeah I've no idea how it's going to work out and the other side of it is when

you're I mean I see these things very much as being tools in the same way that a camera is a tool it's still to do with the creative impulses of the person using them in the same way that you know like images made in Photoshop 15 years ago were kind of frowned upon by OG photographers and now it's very much seen as you know it's understood that not all fine art photographs are completely photo realistic and indeed you know even if you go back to

people like Ansel Adams you know the prints that he made there was so much that went into the print process in the dark room that you know that's a creative process using using tools and experience and knowledge and all the rest of it and I think the same is true of these AI tools that really the the it's still an artistic process and it's still a tool so therefore if you're creating things using AI tools you should own the works that

you're creating at the moment it's a bit of a gray area really you know I want to start monetizing my work both as NFTs and as prints but at the moment it seems to be a bit of a gray area really in terms of whether I own the copyright to the images that I'm creating in the same way that I would do if I was making them with a camera. I'm sure there's there should be ways around it for instance imagine if you create the image on AI you print it out and you take a picture of it that becomes yours because

you take in a picture of a picture and automatically that is your your your your intellectual property or what not yeah I mean I think that's very true and that's something I've thought about I mean the images that mid-journey producers are quite small they're like a just over a thousand pixels on the long edge so I'm going to be using upscaling software to make them bigger and then I've actually been thinking about returning to the analog process so you can you can submit digital files to be printed using silver chemistry technology so in other

words you can actually make an analog print from a digital file already so that's since I'm predominantly using black and white I think that's what I'll be doing is making effectively analog black and white prints from the from the digital upscaled files that are made by AI so there's a nice kind of circularity to that in terms of returning to more traditional processes and techniques imagine if even not not a hundred years but imagine if you set this to a photographer 20 years 10 years ago you would have said like in 10 years time we'll

be producing AI generated images to then print them as 35 millimeter of film people will go berserk well I mean I think even a year ago people wouldn't have understood I mean you know it's the same as with I mean people are more familiar with chat GPT in terms of what's happened with chat bots over the last year but a similar thing that's happened with what they call generative AI image tools can you show us a little example of how it works sure

so this is my mid-journey account and this is where it stores all of the images that I've created and I mean it's very much like looking through a series of negatives to try and find the right one what I do is I run the prompt numerous times so you'll get near duplicates but but some of them are more photo realistic than others when you when you get down into the into the granular detail so it's a question of a kind of a curatorial

process that you go through to find the images that you like so let me walk through the whole process so these were actually I mean I find lots of different ways of trying to get interesting images this here is an interesting experiment so this was version one of mid-journey so this is just read out the prompt to you first which is I think it's at the bottom there eye contrast black and white portrait of Albert Einstein Rembrandt lighting dramatic lighting

like her M11 Sumi Lux 35 mil f 5.6 clean sharp focus lovely bokeh magical mood film photography and then this also had version one at the end so this was the original that's that would would have been the return at the beginning of mid-journey like in whenever it was last November or December or whenever the tool was made public so that was version one that was version two that was version three that was version four which now looks like Einstein

but it also it looks like a kind of a depiction of Einstein rather than a photograph of Einstein when you look closely here is version five so you know it's now very close to looking like an original portrait of Albert Einstein and you know this is all in less than a year so I think the point really is that the improvements that these AIs are making up on an exponential

kind of scale so if you can imagine in 12 months time it will be able to say hello to it it will it will be completely photorealistic at the moment the AI interestingly has problems with hands so it will the hands will be mangled up sometimes or you'll get six fingers or and it reminds me of when you go back into the history of art and you look at the Renaissance and some of these painters they would all start learning their trade with anatomical sketches

so for instance you'll see lots of hands because hands are very hard to draw so you'll see lots of sketches of hands or of muscle definition and things like that these are all the things that AI also needs to be trained on in exactly the same way that an artist learning their trade would be trained on so I found that a sort of very interesting that hands are very difficult for not only for humans to draw but also for AI and I guess we could actually

go to discord so this is this is the discord environment and you'll see that lots of people are trying various prompts here for their could be trying creating an image yeah so so you put in slash imagine and then the prompt comes up and then you're away so what should we create a famous person a famous photographer well I mean it's very hit and miss with famous

people because it depends on the you know I tend not to not to go for famous people it knows very it knows who Elon Musk is very well you know it doesn't know it doesn't you know I've tried you know various different people like Gary Gensler who's the chairman of the SEC and there are lots of images online of him but not enough for the AI to have got a very clear picture of who that person is so famous people is is is quite if you would add someone like Martin Parr or any other photographer that or you mean in the style

of his in the style and maybe him a portrait of a photographer as you showed us a while but Einstein well Einstein is is is is you know there'll be lots of portraits of Einstein online for it to look at so you know I mean I mean we can try but it's an experiment that's likely not to yield who's the most famous photographer that everyone knows about Steve McCurry person but person doesn't have images of him doesn't he yeah I think it might be

a bit of a dead end and Picasso yeah okay okay all right so let's try black and white portrait of Picasso in the style of in his studio painting lighting let's just try like a M10R since I know that tends to give quite nice results

and then you can add various things like the aspect ratio that you want so then you put AR which stands for aspect ratio and then a space and then you could have 16 by 9 or 3 by 4 I tend to use 3 by 4 for portraits but you could have 4 by 3 if you wanted a landscape image so 4 by 3 let's see what that comes up with so it submits it to the server and then you're in a queue basically where you're time sharing

the the processing this is a subscription incidentally which is the basic package which I have is about $30 a month so it's not cheap but you know on the other hand it's a you know it's a revolutionary kind of tool so I think it's worth it's worth the money so yeah so yeah so you now see the image kind of appearing and it it does for iterations you'll see not

all of these are black and white so there's a coherence thing sometimes so it becomes a very curatorial process but you can see now it's 62% complete and you see the image appearing like strangely in the same kind of way that you would see an image appearing in a in a dish in a in a dark room you know it's a similar kind of like weird process of seeing it emerge and okay so it's finished now so we're on a low bandwidth connection

here normally that would be instantaneous Jesus Christ so I mean you can see it's not it's a likeness of Picasso but it's not it's not an exact that looks like him yeah like that really looks like him maybe the first image doesn't look as much but yeah this one yeah yeah and you can see here look his hand is all mangled up it is true yeah and that doesn't look like Picasso at all so it's hit and miss so anyway you've got you've you've

said you like that one so then what you would do is you would upscale that it's number two so it's left left to right so top left top right bottom left bottom right this one two three four so if you want to upscale that you go upscale to you too and you might also decide that you want to run so now it's upscaled that so that the and that's a 1000 pixel image on the long edge so so perfectly usable for web perhaps less so for print in terms of

print resolution but and then what you could also do is you could vary this you could say very subtle or very strong you can zoom out by 1.5 x or 2x or you or a custom zoom somewhere between the two you can turn it into a square so there are various different functions that you can then run on on on the version 5.2 so this is waiting to start now so what it will do is it will take this as the original image and now it will create four derivative

versions of that image that are subtly different so say for example if that if if the hand wasn't perfect on that you could run another iteration and you know by the law of averages one of the hand one of the versions would have a hand that was usable or something so here you have four and you'll see that actually he's wearing slightly different outfits and the painting that he's making is slightly different and there's a statue in the background here the statues are different there's no statue here so you've got very subtly different and also

he looks slightly different in each version. I still think number two is the best one again. Yeah so I mean that's mind blowing. Yeah so you can see the potential anyway and when you think that these this is only going to get more and more coherent you know every month until it's indistinguishable from Picasso you know you wouldn't be able to know whether that was a real photo that was taken of Picasso or or an AI generated one.

So where do you think this whole AI concept is going to go in the future what do you think people be able to do and how are we going to be able to use these photographs because I don't see I don't see I mean of course there's a massive difference between being a photographer capturing the light and there's a massive difference between being a photographer and creating the light. Yeah I mean it's I've been so kind of engrossed in just the process of learning and iterating and seeing what I can come up with and experimenting with light

and white and why I've started with with black and white is because you know well you'll know as a photographer that you kind of if you shoot lots of black and white you tend to think differently about light and shade and contrast and the kind of images that you're creating you know some black and white images will work only in black and white and some images will work only in color you know. So I've started with black and white as a kind

of a first principles thing to try and master the art of black and white AI photography before I start getting involved in the in the realm of color and I did the same with with analog photography as well. So that is to say I haven't really had I haven't really taken much time to reflect on on the bigger picture stuff I mean obviously it means I did an MA in photography and I learned a new word when I was there which was indexicality

and this is to do with the photo being an index of something that it's representing so in other words if you take a picture of a vase then the photo becomes an index of the vase in other words it says that this is a photo of an object in the real world that's a vase so it's indexical of the vase and pretty much since the beginning of photography that's been the basic assumption is that a photograph is is a record of something that exists or existed in the real world. Now obviously people like Stalin airbrushed Trotsky out

of out of images so there's you know there's been some kind of contention about the fact that a photograph represents a real scenario in in the three-dimensional world but to all intents and purposes you know the way that we use camera phones for example we're used to photographs being records of events or objects that that exist in the real world and I think with the dawn of this AI we can no longer and that's across deep fakes and

all kinds of other areas but basically that that that thing that has held photography in place is is now dead basically photographs can no longer be indicative of something that happened in the real world. Wow but if you think about it that is true like with these tools when you're unable to recognize whether it's a true photograph or an AI generated image I wouldn't want to tell a story about this is how was inspired to get you on the

podcast because I know that you do AI photography but I was never kind of hooked on it until I saw an article about I don't know the names I don't know who he is where he's from or anything until this gentleman submitted his work for Sony photography awards. Right yeah I saw that story too. And he won the competition when AI generated photograph. Yeah he he he did some post-processing so there was part part of that was Photoshop as well but the

raw material that he used to create the image was AI generated and he didn't accept the award he went to collect it I mean on one level it was a bit of a PR stunt but he so he went to he went to the whatever it was the the ceremony and basically stood up and said that I created this with AI and AI is not photography therefore I don't accept the award and you're going to win one of the most prestigious photography awards because I

think Sony photography awards are the most prestigious ones yeah and you refuse it because it's yeah yeah yeah so that kind of says it all really doesn't it when when AI generated images start winning photography awards and we're we're less than a year in you know we're less than a year in and cut thank you very much right this was amazing honestly this was amazing and I hope that people who will listen to this they will get something out

of it because this is so new honestly this is so new that they can't even keep up with it yeah man last year an image that is generated but AI wins a photography award what's next yeah I mean I think I mean you know from my personal perspective I think I'm going to try and ride this wave because I think it's I think it's such an important development that there are going to be a number of of early AI artists who will be recognized in

the you know in the same way that people who who were early proponents of cubism or so you know it's it's a genre in its own right and therefore people who crack it early on will will become you know for one of the better expression famous artists so I hope so because it's a form of art I mean I personally don't see this as photography because it's not because you're not capturing anything but it's imagery yeah imagery and photography by the end of it in a digital world the same thing yeah I mean I think the difference is the photo

realism isn't it I mean I've seen people sketching you know you see them on Instagram people who can sketch things and they do the eyes so amazingly that you could barely tell the difference between it being a photo realistic eye and an eye that was drawn with the human hand but sometimes it just looks the same thing yeah if you put them in front of you it's a okay this painting looks more of a photograph and a photograph looks of a photo yeah yeah yeah so it's a similar thing and I agree with you it's not it's not photography

but but there's certainly a blurring of the boundaries you're using quite a lot of the same skills a knowledge of light and shade of lighting of camera angles which lenses to use you know all of that kind of stuff potentially at least comes into play in the same way you know and it's not just going to stop with stills either this is going to move into video so that you could actually write a script like a director would write a script with you know storyboarding and explaining the scene and then the dialogue and all the

rest of it and you know I'd say probably within 10 years there'll be there'll be full length feature films made by AI I'd love to see that I would really love to see that because you can make a movie out of anything in any style at any point yeah with no budget but no production budget whatsoever yeah that would change the world and it really would yeah I mean I've already heard people say that about some of my photos like to take that picture would have required makeup someone organizing a travel budget you know a full itinerary yeah

model model releases legal all of this stuff and basically you just click to button and there it is you know if you think about it this this this could be so revolutionary like it really could be from every point of view yeah from every point of view yeah and as you said earlier I think I completely agree that to generate an image like you do it's not just saying black and white portrait of a lady with cool lighting in the background off yeah like I am 11 or what what thought with the with the 56 millimeter 1.2 yeah people

don't know this like yeah technical you are the better image you're going to produce as a photographer yeah totally yeah yeah yeah and you know I'm still learning but it's definitely a process of refining prompts you know they're already calling people prompt engineers you know there's a so you know it's there's a new classification of job called prompt engineer now I mean you have to know a python and a few other things but you know there are there are people now earning three and four hundred grand a year specializing in being a prompt

engineer so that they know how to interrogate chat GPT for in order to get amazing results much better than the average user stumbling into chat GPT would be able to get because they know how to you know optimize the text that they're putting in and to they know all the tricks of the trade in order to generate the best results from the software and I think you know the set on a smaller scale the same is true of the the the image side of things you know I'll definitely explore this image making this image generating for photography

as a business because why not yeah why not yeah well thank you again Brian this was amazing pleasure man

Open the transcript file

Hosts and Guests

Nobody is credited on this episode yet.

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. Dr Oliver Zolman - Can We Reverse Ageing? The Future of Human Longevity

    Nov 16, 202625 minAudio

    What if we could slow down, stop, or even reverse the ageing process? And what would an extra 50 years of healthy life mean for humanity? In this episode of Everything Cambridge, we speak with Dr Oliver Zolman, a...

    Credit the people on this episode

  2. Elisabetta - From a Life of Travel to Art, Identity & Sculptural Jewellery in Bali

    Nov 12, 202633 minAudio

    What happens when you follow your curiosity, leave everything familiar behind, and allow the world to shape who you become? In this episode of Everything Cambridge, we sit down with Elisabetta, an artist, dancer,...

    Credit the people on this episode

  3. Dinky Doors – The Secret Artists Bringing Magic to the Streets of Cambridge!

    Nov 12, 202644 minAudio

    What if the smallest things could bring the biggest smiles? 🚪✨ In this episode of Everything Cambridge, we meet the mysterious creative minds behind Dinky Doors, the miniature hidden artworks that have transformed the...

    Credit the people on this episode

  4. Tez Ilyas - From Science Degree to Stand up Comedy

    Oct 7, 202649 minAudio

    In this episode of The Cambridge Chronicles Podcast, Sam Melvin sits down with comedian Tez Ilyas for a funny, honest and uplifting conversation about comedy, ambition, failure and finding your own voice. Tez talks...

    Credit the people on this episode

  5. Paul Butterworth -The Full Monty, 50 Years Acting & Surviving the Industry

    Jan 26, 202644 minAudio

    Actor Paul Butterworth reflects on nearly 50 years in film and TV, from The Full Monty to British soaps, sharing honest insights on surviving as a working actor, treating small roles like leads, being cut from films,...

    Credit the people on this episode