Someone pointed out to me that ChatGPT can create a Film Simulation Recipe that replicates the look of any photo that you upload. Just share a picture with AI, and it will analyze the look and produce a Recipe that matches. But can it really do that? And, if so, is it accurate?
So I gave it a try. I uploaded a picture and asked AI to replicate the aesthetic of it with a Fujifilm Recipe. Sure enough, it produced one. I also asked ChatGPT to tell me which Fuji X Weekly Recipes are the closest. Let’s examine the results.
Test 1
The top picture is the camera-made JPEG captured on my Fujifilm GFX100S II using the Nostalgic Film Recipe (a Film Dial Recipe). I uploaded it to ChatGPT and asked it to make a Recipe based on that image. The second picture was reprocessed in-camera using the settings that AI provided. Quite obviously, they are pretty far apart and are not a good match.
Some of the settings are the same. Both pictures use DR400. Both are Grain Weak (although ChatGPT chose Grain size Large). Both have Color Chrome Effect Strong and Color Chrome FX Blue Weak. Noise Reduction is -4 on both. The White Balance Shift is +2 Red & -4 Blue. All of the other settings were different—for example, the Nostalgic Film Recipe calls for the Nostalgic Neg. film simulation, while AI chose Classic Chrome. The Recipe made by AI isn’t too dissimilar to Kodachrome II, although not an exact match. I asked ChatGPT to tell me which Fuji X Weekly Recipes were the closest, and it suggested Kodak Ektar 100 and Kodak Gold 200.
Test 2
The top picture is the camera-made JPEG captured on my Fujifilm GFX100S II using the Fujicolor PRO 160C Warm Recipe (which is also a Film Dial Recipe). The bottom picture was reprocessed in-camera using the settings that ChatGPT provided. They are far apart, and are not an especially good match, although it is slightly better than the first test.
The Fujicolor PRO 160C Warm Recipe uses the Reala Ace film simulation, while ChatGPT chose Eterna. Most of the settings are divergent. I don’t have a Recipe that’s especially close to the one made by ChatGPT, but Timeless Negative is probably the most similar. The Fuji X Weekly Recipes that AI suggested are the closest are CineStill 800T, Kodak Ektachrome 100SW, and Kodak Vision3 250D.
Test 3
The top picture is the bottom image from Test 1. I asked ChatGPT to make a Recipe that matches it, to see if it would suggest the same settings that I used to make it (which is the Recipe AI had made just a few minutes earlier). The Recipe that it came up with was not the same, and also not a great match. Instead of Classic Chrome, it wen’t with Classic Negative. While only somewhat similar, Fujicolor Blue is probably my closest Recipe. The Fuji X Weekly Recipes that AI thought were the closest are Kodak Portra 400 and Kodak Portra 160.
Test 4
I re-uploaded the first photo from Test 1, and asked the same question to see if it would suggest the same settings, or if it would change them. Unsurprisingly, it was completely different, this time choosing Astia for the film simulation (I didn’t bother to reprocess). It suggested that my Kodak Portra 400 Recipe would be the closest match.
Conclusion
I’ve said before that ChatGPT is a neat little card trick, but once you know how it works, it isn’t nearly as impressive. I’ve also said that it will only get better with time. Both of those statements still seem to be true. While AI can analyze a picture and produce some Fujifilm settings to try—which is amazing—it isn’t very accurate. It can also suggest Fuji X Weekly Recipes that might be similar, and it is a little better at that, but still not especially good.
I have zero doubts that with enough time—and it may only be a couple of years—ChatGPT will be able to analyze a photo and produce a Fujifilm Recipe that is pretty similar. It might be even closer to suggesting an existing Recipe that isn’t far off. It’s not there yet, though.
I’m certain that ChatGPT was trained on my website. When asked directly, it says that it wasn’t, but states that it was trained on various photography blogs and forums. When I asked it to name some specific ones, the very first it listed was Fuji X Weekly. So, yeah, AI did some of its learning from my website, which of course I did not consent to nor was I compensated for. I believe that this training is ongoing, and it probably will even learn from this article that you are reading right now. This is wrong, and eventually the courts will likely determine that it is completely illegal; however, by then, the damage is done, and there’s nothing that can be done.
With all of that said, I had a conversation with a friend of mine who is super intelligent (he’s a scientist) about AI and ChatGPT. He said that he is not concerned about being replaced by AI, but he is concerned about being replaced by someone who knows how to use it better than he does. I don’t think it is prudent to ignore ChatGPT completely. While undoubtably ethically sketchy, it is a tool that can sometimes be used effectively, if you understand its strengths and weaknesses. I’m not sure what that means for me specifically, other than it is good to test it now and again to see how far it has or hasn’t come. It definitely has made some progress since the last time I tested, but not a lot.
Interesting… You should replicate the same test with DeepSeek!
I don’t believe deepseek has image recognition functionality yet. Gemini does, and could be worth a try, although really these AI tools are just bullshit spitters!
I just tried out Gemini. It’s similar to ChatGPT but a little worse for this. For example, it suggested Grain of -2, and also omitted some settings. It reminds me of how ChatGPT was a year ago regarding Recipes.
When I asked Gemini how it knows about Film Simulation Recipes, it provided me with a few different sources, including this article:
https://www.shopmoment.com/reviews/why-shooting-jpegs-is-so-popular-10-fujifilm-simulation-recipes?srsltid=AfmBOoqzHIhpl0hlB1TcafEkUC01Ik2tYTG_hP455elDyVZvHAk_1FYU
I asked specifically if it was trained on Fuji X Weekly, and it responded with, “It’s possible.” 🤣
I haven’t tried DeepSeek. There are a few things I understand about it:
1. You cannot register for it right now in the USA, because Google, Yahoo, etc., are blocking emails from DeepSeek, preventing registration. I’m sure this is at the direction of the government.
2. The terms of service for DeepSeek is extremely sketchy. Basically, you agree to give them access to your phone or computer, including keystrokes, which can be used to determine passwords and pin codes. The intention is probably to further train it off of the contents of user’s devices, but it could certainly be used nefariously. I hope they reel this back a whole bunch, because it is quite risky as it stands now.
3. It was trained largely on ChatGPT (oh, the Karma…), so I suspect that whatever it gives will be quite similar to ChatGPT.
I tested it on the look of SX-70 Polaroid film.
And every time I asked again it gave me completely different values.
So if you ask me with words it’s pointless and clueless.
Maybe uploading an actual SX-70 Polaroid is a different thing.
Another thing I see as a problem with AI is that it starts slowly to copy itself because there’s so much AI trash on the web already.
A lovely inbred is awaiting us 😂
I’ve not been able to get it to give me the same answer twice, which means that it’s just making it up as it goes along (which it does). That inconsistency alone makes it unreliable. I also agree on the inbred comment. The more you ask, the weirder the answers get.
Interesting Ritchie for this second trial.
I’m very glad though it missed again!
But as you stated it is only a matter of time.
Being an IT guy, my value is that IT should serve and help human doing it’s tasks and simplify them. Not the other round or enslave.
Unfortunately the path choosen especially for IA, is a race, like the nuclear race between superpowers. Only here it private companies owning it…
And what great purpose after all, except making more money ?
This time as technical revolution in contrario of the industrial ones, it will destroy jobs and it will replacing them. Where your scientific friend is absolutely right on the spot is that (the oned at the top of the pyramid that is) will be replaced by someone who knows how to use it better than he does.
My issues with IA are several points.
1) As your clearly stated : it got trained on your data without your explicit authorisation or even asking for it. Consuming your data you created throught a long process, trials and build up of expertise. Which should be remunarated at is correct value.
New jobs about IA and copyright issues are emerging. Huge and very sensitive subject, impacting directly photographers like us.
But you are right also, the arm is already done : it grab and trained for free on your hard human brain work.
2) Quality of data. In order to give proper and accurate results from its statical computation, data being fed to IA must be ‘clean’, accurate, scientifically proven, etc.
For private companies to start using IA, this what we (IT guys) says every time so the use of IA can relevant in the company field of activity
Unfortunately the web and dark web especially and social media are more than full of heavy junks… And by stopping moderation, accurate and none partial research, this will not bring free speech at the end but more fake news and so bad data to be processed.
3) Generative IA like ChatGPT are running of content juice. Aka the free data found on the web had been nearly consumed, forcing IA companies to retrain their IA on existing data and they are trying on the hard way now to find other / new content without willing to pay for it (as it is a huge cost for them this reducing the profits made), thus aggravating problems 1 and 2 in expential ways.
4) Probably the main one : energy consomption !
The compute power needed for generative IA is immense and needs to grow expentialy, following the Moore law pretty much closely or even surpassing it.
This means that data centers holding it have to grow the same way. This is demanding tremendous energy production and consumption (electricity for the servers, storage, network, cooling systems which are ecological friendly, etc). Google has already set to put small nuclear powerplant nearby its data centers so to sustain it’s electrical needs (constant supply, independence vs energy companies and more so the actual huge financial market variations on energy markets) so to sustain it’s forecasted growth. Microsoft is submerging smal data centers under the sea, etc.
This going completely against climate change needs where we should be reducing our energy consumption in order to win against. Last GIEC report is more than explicit!
Leading only to aggravate the situation in every way (energy consumption rising instead of decreasing, global worming not slowing down, huge dicothomy between the ones having it, and the ones deprived of it, etc).
We are going on the very wrong path here. I will not say to fear IA, but in fact stop using it!
Or at least ask oneself the question : do I really need itto use it?
Yeah, using it to understand science problems, corrolate huge data together, data mining, weather prediction, image enhancement / reconstruction, complex nano surgery, IA there serve terrificly its purpose.
But do I need to generate a simple film, an image upon my simple desire of the moment for simple fun ? For what ? Each prompt made by billions of users cost a tremendous energy consumption behind!
One prompt is fine, billions, nope!
At the time we are living in, with climat change will come hard on us, if we are not careful ? As an IT guy, and father of 3 young adults, I would like that my children and grandchildren be able to live in a world where I had the chance to live in.
So my answer is definitely NO !
We as users and consumers, and data producers (text, image , etc) we should say NO !
This is the only way to stop this pure commercial race to stop or at least slow down.
A french comic said in the 80′ :
“If only people would stop bying goods, companies would stop producing / sell them”. Well while it was funny at the time, he was right on the spot concerning IA.
This was is like the other emerging tools : it is for the moment a fashionable tool. Every body wants to use it.
But we should cool down and take a step backward and think very hard.
A year 0 should be made on a global scale on this subject like critical other ones.
If not, well, we are in a speed car with real driver on the path to hit at high speed a concrete wall…
I do hope we will be smart enough to stop / push the breaks before the crash and not asking a generative IA what to do…
People don’t really talk about the energy consumption. I remember reading a couple years ago that AI is expected to use approximately 25% of the world’s energy consumption by some such future date (maybe 2030? I don’t remember). That was shocking to me, but it doesn’t really get discussed much, and not many seem to be bothered by it. Crazy.
Indeed crazy! And haven been an IT infrastructure consultant for an energy company for quite some time, this IA energy consumption is one of the big challenges to tackle for the coming years.
Your numbers are already outmatched by the way as IA is growing even faster now…
Hence why I brought here the subject here as photographers who can use Generative IA for their work…
The less AI in the world, the better, for a multitude of reasons. Sadly, there’s just not enough concern for the consequences (and there are a lot of large consequences). Unfortunately, I think we’re quickly approaching the point of no return, and it’s ramping up and not slowing down. It might be ok if everyone exercised moderation, but that seems against human nature 🤣. What I think it will be in just a few years is either you are fully onboard or you are left behind, there just won’t be any escaping. That could prove to be a brave new world or a dystopian nightmare (which might even be specific to the person or region)—let’s hope for the former and not the latter.
I fully concur with you!
And you point 2 critical items nowadays:
1) moderation : in energy consumption, consumerisation, etc.
We are speeding up like IA, not cooling it down.
But also likewise moderation in how we exchange / talk to each other in forums, blogs, social media, etc. No more moderation there. No cross-checking facts, etc. This is leading too extremes, bloody fake news, all what see nowadays and even more…
2) Leading to your second crutial point : a brave new world (if people come back to their sense) or a dystopian nightmare as you wrote very well which many leaders over the world (elected or not) are preparing us unfortunately.
Capture One does this as from the latest update. At least for post work in capture one.
Just drag and drop or copy any image into the tool – and it will stay as a kind of preset to paste onto other images. Best used with several images in the tool as no image will have the same contours, shadows etc. Like your examples, the results get you only so far and then you have to fine tune.
But to use ChatGPT and all such engines effectively, you need to correct it’s errors. It’s a tool for experts, not an expert tool. A layperson might scratch the surface and get excited, like someone that uses Photoshop to crop only. But the power lies below the surface.
Unfortunately these tools will create a new generation not only that feels entitled to everything instantly – as is already the case – but without any effort, and without any consequence on their quality of life (work, environment, health).
Oh, interesting. I would be curious to test how well the Capture One feature works. A test could be to make two photographs of the same scene: one a SOOC JPEG with a Recipe, and one a RAW file where the camera settings were factory default. Then see how well Capture One can match the JPEG aesthetic to the RAW.
An interesting poat again, thanks for that.
I have a question though: what is meant by “Film Dial Recipe”? Is this specific to the GFX or am I missing something else?
The Film Dial Recipe, sometimes referred to as the Universal Recipe (someone on YouTube called it “the Recipe of all Recipes”), was made for the Fujifilm X-T50. It’s also great for the X-M5, which also has the Film Dial. Interestingly, it can be used on any newer Fujifilm camera, and doesn’t actually require it to have a Film Dial.
You can find it here:
https://fujixweekly.com/2024/05/16/fujifilm-x-t50-film-dial-settings-14-new-film-simulation-recipes-yes-14/
Okay and thanks Ritchie so, if I get it right, I understand, why I was not aware of the Film Dial Recipe. I just own am X100VI amd owned an X100V for a couple of months.
You can program the front control ring on the X100VI to switch film simulations (instead of Digital Teleconverter), and use the Film Dial Recipe in conjunction with that.