AI at IVAE Studios only culls, sorts and backs up files. It never generates skies, faces, bodies or guests, and nothing enters or leaves the frame beyond genuinely temporary details like a blemish or a stray hair. Every image in the portfolio is a real client on a real day, ahead of the EU transparency rules applicable from 2 August 2026.
You can no longer tell, from a photograph on a website, whether the sky in it existed. That is a new problem and it is a bigger one for photographers than for clients, because our entire business rests on a promise that the thing in the frame happened. Consumer trust has moved fast in response. Gartner has found that around half of US consumers prefer brands that avoid generative AI in customer-facing content, and the European Union's transparency rules for synthetic media become applicable on 2 August 2026. So rather than wait to be asked, here is IVAE Studios' written policy: exactly where we use machine learning, exactly what we will never do to a photograph of you, where the honest grey area sits, and how to check any photographer's portfolio for yourself.
What We Use AI For: Culling, Backup, Sorting
We come home from a wedding with several thousand frames. Sorting them by hand used to take a full day before any editing started. Software now groups near identical frames, flags closed eyes and missed focus, orders a sequence by capture time and matches the second camera to the first. That is machine learning, we use it, and it changes nothing whatsoever about the photograph. It is clerical work.
And then the grey area, which we would rather name than hide. Modern masking tools in Lightroom and Capture One use machine learning to find a sky, a subject, a background or a face so you can adjust each separately. When we lift the shadows on your faces without lifting the sand, that selection was made by a model. Our position is that this is selection, not generation: it decides where an adjustment lands, it does not invent anything that was not photographed. We are comfortable with it and we think you should know it is there.
What Is Banned: Generated Skies, Faces, Guests, Bodies
This list is not aspirational. It is the rule the studio works to, and it applies to every frame that leaves here.
Sky replacement is banned, whether generative or pulled from a library. If the sky was flat and grey that evening, your photographs have a flat grey sky, and we will have used position, timing and exposure to make that work rather than swapping it afterwards.
Generative fill and generative expand are banned. We do not invent beach, sea, foliage, floor or wall that was not in the frame, and we do not widen a crop by asking software to imagine the edges. If the composition needed more room, we needed to stand further back on the day.
Faces are never generated, swapped or composited. No taking the smiling head from frame 12 and putting it on the body from frame 14. That includes group photographs, which is where the practice is most common and most defended. If somebody blinked, we deliver the frame where they did not blink.
Nobody is added and nobody is removed. If a guest walked into the aisle during your ceremony, that guest is in your ceremony photographs. If a stranger was on the beach behind you at sunset, they were on the beach. Removing a person is inventing a version of the day that did not happen.
Bodies are never reshaped. No slimming, no lengthening, no jawline work, no waist edits. We do not offer it as an upsell and we do not do it quietly. The only exception is if you specifically ask us to, and even then we will tell you what we think before we agree.
And nothing on our side is synthetic either. Every photograph on this website, in our portfolio and on our Instagram is a photograph of real clients on a real day on this coast. We do not use AI generated images to advertise a service that is supposed to produce real ones. That should be obvious and, in 2026, it is not.
Retouching Versus Generation: Where the Line Sits
The test we apply is a single question: does this edit describe something that existed, or invent something that did not?
On the describing side sit all the things photographers have done since the darkroom. Exposure, contrast, white balance, colour grading, dodging and burning, lens and perspective correction, noise reduction, sharpening, dust spot removal, cropping and straightening. Also small removals of things that were genuinely temporary: a blemish that was not there last week, a stray hair across a face, a piece of litter on the sand, a sticker on the sole of a shoe, a price tag, a hair tie on a wrist in a formal portrait.
On the inventing side sits everything in the banned list above.
Now the honest middle, because a policy that pretends the line is clean is not worth much. Modern healing and removal brushes are increasingly generative under the hood. When we remove a water bottle from a table, some software fills the resulting hole by generating plausible tablecloth. Our practice is to use the non generative content aware option where the software still offers one, to keep removals small and structural, and where neither is possible, to simply leave the bottle. A bottle on a table is not a problem. A generated tablecloth is a small lie, and small lies scale.
On skin the rule is temporary versus permanent. A spot goes because it will be gone in a fortnight anyway. A scar, a birthmark, a freckle, a line, a tattoo, a gap between teeth: those stay, because they are you, unless you ask us specifically and we will talk about it first.
Content Credentials and Camera-Level Provenance
There is now a technical answer to the question of whether an image is a photograph, and it is worth understanding because it is going to become normal. Content Credentials, built on the C2PA standard from the Coalition for Content Provenance and Authenticity, attach a cryptographically signed record of where a file came from and what was done to it afterwards.
Cameras have started signing at the moment of capture. Leica shipped the first body with it built in, the M11-P, and Nikon, Sony and Canon have rolled out or announced firmware for their professional bodies. Adobe supports reading and writing the credentials in Lightroom and Photoshop, so an editing history can travel with the file.
The honest limits matter as much as the promise. The metadata only survives if every step of the pipeline preserves it, and most social platforms strip, flatten or partially display it. A screenshot destroys it entirely. So provenance today is a real and useful signal, and it is not yet a chain you can rely on from camera to Instagram post.
The EU Labelling Rules That Start in August 2026
The European Union's AI Act phases in over several years, and the transparency provisions in Article 50 become applicable on 2 August 2026. In plain terms, anyone deploying an AI system that generates or manipulates image, audio or video content constituting a deepfake has to disclose that the content has been artificially generated or manipulated.
There is a carve out, and it is the one photographers should read closely. Edits that are assistive, standard, and do not substantially alter the input are treated differently. Colour grading and exposure are clearly on the safe side of that. A generated sky over a real ceremony, or a guest removed from a real aisle, is on a plain reading a different thing. We are not lawyers and this is not legal advice, and the practical guidance will be argued over for years.
We are a Mexican studio, so the EU is not our regulator. It still matters to us for two reasons. A meaningful share of the couples we photograph are European, and the photographs go home with them and get posted there. And more simply, a rule that pushes an industry toward saying what it actually did is a rule we already comply with, because we do not do the thing.
How to Audit Any Photographer's Portfolio for AI
You do not need software for this. Nine times out of ten it is visible if you know where to look, and the checks take about ten minutes.
Extremities. Hands, fingers, ears, teeth, jewellery, watch faces, buckles, lace. Generative models still stumble on small repeated structures. Count fingers, check that both earrings match, look at where a ring sits.
Repetition. Zoom into sand, foliage, brick, tile, and crowd faces. Generated fill repeats texture in a way real optics never do, and once you see a duplicated clump of seaweed you cannot unsee it.
Light logic. Every shadow in a frame should point the same way, and the direction should match the sky. Reflections should contain what is actually in front of them. Water should reflect the sky it is under.
The same sky twice. Scroll a whole portfolio looking for identical cloud formations behind different couples on different dates in different places. This is the single fastest way to catch library sky replacement.
One full wedding, not a highlight reel. Ask to see a complete gallery from one day. Fabricated or heavily manipulated frames rarely survive next to their own neighbours in the same sequence, because the light, the crowd and the background have to stay consistent minute by minute.
Ask for a RAW. Any working photographer can send you an unprocessed original from a portfolio frame in a couple of minutes. An inability or an unwillingness to do that tells you what you need to know.
Ask the question in words. Do you replace skies. Do you composite faces in group photographs. Are the images on your site your own work from real clients. Vague answers are answers.
Why the Imperfections Stay in Your Gallery
The wind on this coast is constant. Hair goes across faces, dresses wrap around legs, sand moves. Somebody in the family group is always mid blink. The sky in October does what October does.
Every one of those is fixable now, cheaply, in software, and we are telling you that we are choosing not to. The reason is not purity. It is that the frames couples come back to ten and twenty years later are almost never the perfect ones. They are the ones with evidence in them: the specific wind of that specific afternoon, a laugh at something a guest said, a sky that was doing something strange. A corrected image is a picture of a generic evening. An uncorrected one is a picture of yours.
What we do instead of fixing things afterwards is get them right in the moment. Golden hour, only, which is a scheduling decision rather than an editing one. Position rather than plugins. Scouting a resort so we know which wall is lit at 6:15 in March. That is the whole method and it is why the policy above costs us nothing.
Common Questions About AI and Your Photos
Do you use AI at all?
Yes, for workflow, and we would rather say so than pretend otherwise. Machine learning helps us cull thousands of frames, group faces so we can find every photograph of a specific guest, verify backups and transcribe our own notes. Masking tools in Lightroom also use machine learning to select a sky or a subject so an adjustment lands in the right place. None of that invents pixels. Generative editing of your photographs is banned.
Do you replace skies?
Never. If the sky was flat and grey that evening, your gallery has a flat grey sky and we solved the picture with position, timing and exposure instead. Sky replacement is the single most common generative edit in wedding photography and it is the first thing on our banned list, because it changes what the day looked like.
Will you slim or reshape us?
No. No slimming, no body reshaping, no jawline edits, and it is not offered as an upsell. We retouch temporary things such as a spot, a stray hair across a face, a wardrobe tag or a piece of litter on the sand. Permanent things such as scars, birthmarks, freckles and lines stay unless you specifically ask us otherwise.
Can we see the original camera file?
Yes. Ask for any frame in your gallery and we will send the original camera file. We keep them, and being able to produce one on request is the simplest possible proof that a photograph is a photograph. It is also the question we suggest you ask any photographer you are considering.
Does the EU AI Act apply to our wedding photos?
We are a Mexican studio, so the EU is not our regulator, and none of this is legal advice. But the Act's transparency provisions in Article 50 become applicable on 2 August 2026 and require disclosure when image, video or audio content has been artificially generated or manipulated to a degree that makes it a deepfake. Standard, assistive edits that do not substantially alter the original are carved out, which covers ordinary colour work. A generated sky over a real ceremony, on a plain reading, is a different matter. Since we do not do generative editing at all, there is nothing for us to disclose.
How can I tell if a photographer's portfolio is AI generated?
Check hands, fingers, ears, teeth and jewellery, where generative models still stumble. Look for repeating textures in sand, foliage, tile or crowd faces. Check that every shadow in a frame points the same way. Scroll for the same cloud formation appearing behind different couples on different dates. Ask to see one complete wedding rather than a highlight reel, because fabricated frames rarely survive next to their neighbours in a sequence. Then just ask directly, in writing, whether any delivered image contains generated content, and keep the answer.
IVAE Studios was founded in Cancún in 2023 by Vianey Díaz, who has more than ten years behind the camera and leads every session personally. Over 500 sessions delivered, a 5.0 Google rating from 47 reviews, bilingual in English and Spanish, golden hour only, and galleries in 1 to 3 business days at 4K in a private gallery. Everything you see in that gallery was photographed.
You can read more about how the studio works on the about page, see the work on couples photography in Mexico, find us on the coast at Cancún photographer, and browse the rest of the guides in the Journal.