AI (Artificial Intelligence) photo editing has been around for a few years now. How have programs which implement Machine Learning and AI to simplify complex editing tasks changed the photography industry? What are the positives and negatives for utilizing these programs in your editing workflow, and what is on the horizon?

The driving forces

If you are a professional landscape photographer, meaning the bulk of your income is derived from your photography, then you understand the whims of the market (your customers and people buying your images) tend to drive photographers toward surrealistic image styles because these often sell over those which look more realistic. An example of this is the trends for the last few years of super saturated (unrealistically vibrant) images, HDR or selective color. All one needs to do is go peruse a local art and wine festival to see what the latest trends are and how they are selling. I visited one recently and according to those selling images there, the oversaturated images are still selling quite well, but what I saw even more of was AI enhanced images with sky replacements, and photos with unrealistic light (time of day) changes. Many looked as if they were done hastily and displayed artifacts and masking issues. It reminded me of the early days of HDR where everyone was cranking the amount sliders up to 11 on everything, generating images that looked like something out of a cheesy post-apocalyptic movie.

The market demands the perfect sky, taken in the perfect light, and AI powered programs with a few clicks can create a reasonable facsimile of a perfect image with little to no retouching knowledge. If these AI programs can really replicate the time of day, season, and sky that you want, what is the incentive for the landscape photographer to keep going back to a location in order to capture an image with the perfect light?

What is AI Photography?

Programs like Luminar AI (and NEO) utilize artificial intelligence to analyze the structure and objects in your image and allow you to change aspects of it with a few sliders or clicks. You can replace the sky, open the shadows, add smart contrast and vibrance quickly and easily by selecting one of the included presets.  Even Photoshop has gotten into the one-click game, their neural filters can do a lot of heavy lifting and get an image well on its way to final. In the latest builds, the new landscape mixer neural filter allows the user to convincingly (at least to the general public) adjust the time of day, time of year and of course replace the sky. In the case of sky replacement, Photoshop yields a multi-layered file with great looking masks that the user can tune to their heart’s content. [note] Neural Filters and sky replacement are only available in the latest versions of photoshop.

What can AI image editing do for an image?

For this example, I chose a five-image panorama of Dream Lake I took the last time I was in Rocky Mountain National Park. We were hiking to Emerald Lake for sunset shots, so this photo was taken at late afternoon. The only thing I have done to this image after stitching the panorama together is sharpen using Topaz Sharpen AI (you can see I have the layer for the landscape mixer turned off in the layers panel.). Yes – I usually do reserve sharpening for the final step in my editing process, but this pano was a bit soft in a few areas to start with.

To activate the landscape mixer, open your image, then select filters > neural filters. If you want the ability to tune your settings later, then select convert for smart filters under the filter menu before heading to neural filters. Turn on landscape mixer under the neural filters panel and you are all set. You can use the sliders to tune your own look or use the presets at the top of the panel.

In the following examples, you can see that I only slide the season slider to full. The amount of the effect applied to the image can be changed with the strength slider. This allows you to tune changes to your liking. In the last example, I added some night and some sunset effect to the Autumn season slider. This two-minute edit netted a file with some odd lighting around the tops of the trees- which could be easily fixed with some clone stamp tool set to darken, or burning under dodge/burn. Overall, I would probably use dodge and burn to lighten up areas of the image to add a bit more interest, but this is certainly a good starting point. In the examples below, this is how the image came out of the filter, I would not consider this a final-worthy edit. The grass generated by photoshop tends to be a bit blurry and lacking in texture since it is creating new objects in your scene. Most of the landscape mixer results would not be ready for prime time in terms of overall quality in my opinion. So you can clearly see the output of the filters I have left them at 100%, in most cases especially with the night filter I would tone the effect down a lot. 

These filters do some amazing things to your images – and they are still in Beta, I am looking forward to seeing the first real releases of these filters.

Downsides of this technology

With AI editing, complex photo manipulation techniques are now accessible to every level of photographer, what is the downside of using these programs for your edits?

• Everything starts to look the same. Users are working off the same preset template and sky files included with the programs. I can pop onto Instagram and recognize a Luminar template or sky image within a couple of minutes. Often users don’t even bother to tune the preset to make their image look unique when compared to others.
• The technology is widely available. Removing a tree or doing complex edits to an image used to be the domain of the photoshop skilled. Now anyone can get decent results from one or two clicks. This devalues the skillset of the accomplished image retoucher and relegates editing to presets.
• Nearly every program does sky replacement with a few clicks, and some of the results can look pretty bad. This is especially true when a midday shot has a sunset sky applied to it, or a scene captured on a cloudy day with diffuse lighting has some crazy sky with harsh lighting added.
• It is easy to Identify AI edited images by looking at the light in the scene. Often users will add a sky with a light source coming from the opposite direction, making the shadows or light cast effect look wrong. Users also often don’t pay attention to the masking in the tops of trees or mountains, leaving a halo of the old sky in the image, this as the same as doing a poor job of cutting out a subject in photoshop. In the case of Luminar AI and NEO, look at the tops of any mountains, they are often darker just below the sky – the mask is darker than it should be, and can easily be identified.

Should you use AI editing?

Editing has always been a philosophical debate in the photographic community. The truth is that every image you see has been edited at least somewhat. If you are shooting RAW, your saturation, shadows highlights, and sharpening will need to be adjusted at a bare minimum, and at that point, the image has been manipulated from the original flat RAW file. Whether to remove a distracting object, go bonkers with the vibrance and saturation or replace a sky all depends on the use for the image, and if you are representing that the image looked like that when you took it. In the case of my example image, it is an ok shot, not great – compositionally it is fine, but lighting and time of day is not optimal for an image I would consider putting in my portfolio (that is why it is being used for this article).

Instagram and other sharing sites users are ravenous for new content and photographers don’t always have the time to hit the landscape trail to take perfect images. Enter digital AI photography and an old image with mediocre lighting can be quickly changed with a few clicks to something that will keep your followers coming back to your profile. Is it untruthful? It really depends. If you are saying you took the photo when it clearly uses a sky file that came with Luminar or Photoshop, then you are going to get caught in your lie. If you are presenting the image as your vision for the scene before you, the public is used to images that have been manipulated, so if it looks pretty, does it matter?  In the case of art and wine festivals, if the oversaturated mess you are offering is selling then who am I to complain about it, the only one who would care about the amount of manipulation will be other photographers and they aren’t your customer.

Seeing both sides

The other side of this coin is the devaluing of the purist landscape photographer’s work. I know landscape photographers that travel to the same location 10, 20 or more times at varying times of day and season to capture the perfect representation of that location. I’m guilty of this myself. I was in one location in Rocky Mountain National Park and loved where the sun set, and how it lit the clouds. I returned a couple times during each season the following year, until I found myself in the same location one year later seeing the sun set in the same location that inspired me to shoot there in the first place. Truth be told, the sun set in the same location, but the sky was a clear blue and did not yield an image as compelling as the original one.

With the glut of photos available on the internet it is easy for a viewer to scroll past an image that may have taken the photographer days, weeks, or years to create – and getting excited over something that was taken midday and manipulated to be something that the market wants.

So, does that mean don’t use these programs if you are a purist, I would say no, they have lots of other effects and time-saving features to offer that can enhance a “truthful” image, and really bring it to the next level. The trap you need to avoid is using the presets, because your images will quickly begin to look like everyone else’s out there.

 Does it matter?

How has AI photo editing affected the industry? Is it a negative that everyone can make pretty images with a few clicks, and can improve the results by learning more about the editing process? Was it a big deal before? No, every time someone saw a beautiful image, they suspected or knew that it had been edited/photoshopped. These programs allow the novice retoucher a chance to have their work stand closer to that of a professional photographer and retoucher.

Instagram and a million other phone apps already offer a variety of different looks for your images, including skin retouching. I don’t think that anyone would agree that the eye treatment and skin retouching on Instagram would put a professional fashion retoucher out of business. It only allows lesser skilled photographers the ability to have their work look better.

Moving forward

Like with any business, the introduction of AI editing will force some aspect of the photography industry to evolve. When digital came to the forefront of the market, companies who created film eventually had a choice, to go out of business complaining about those kids and their digital cameras or evolve their business model. Fuji focused more on professional imaging and cameras, while other companies like illford, who supported film shooters embraced schools and specialty shooters who could not give up film. Think about all the companies that developed film, and supported film shooters, eventually pharmacies added print your own machines instead of offering film developing. In the case of AI editing, retouchers may not see as much business as they did from some segments of the photography industry, just like company’s who focus their business on photoshop training classes. I don’t have the answers, but they will have to face the choice of evolving their business model or failing.

On the horizon

This technology isn’t going to go away, I only see it improving over time. With the introduction of their Neural Filters, Photoshop is already well on their way to offering a variety of one-click style editing options. I would bet they will include sharpening, image sizing and others in coming versions of the program. This would make Photoshop more of an all-in-one solution as users will have the choice to use the one-click functionality or do it the way they have always done their editing.

I see other programs like Luminar NEO continuing to evolve and improve. Masking and selections will become more accurate as their AI algorithms evolve and eventually, they will become indistinguishable from a well-done hand-tuned mask. The issue with this technology will still be that users are working from presets and their work will look similar, so how do we combat this?

Use scenarios

Seasoned photographers and those skilled in retouching can certainly utilize the functionality of these programs to achieve specific tasks that would traditionally take a lot of effort. These programs often do a great job of opening the shadows and adding realistic vibrance to a shot, if users shoot their own sky files, and create their own LUT (lookup tables) you can still achieve unique looking images while enjoying the time saving features of these AI programs.

I have been using Photoshop since version 2, so I can replace a sky, mask in or out an object and color correct an image easily, but just because I can do something doesn’t mean I want to. If I can save myself some time, like adding an Orton effect to an image, using one of these programs I will. This is especially true if the result is close to what I would achieve with my own photoshop actions. I almost always implement sharpening using Topaz sharpen AI, because I can’t see the sharpness of my images when I am editing (I’m almost completely blind).

Technology can’t make you see

These AI photo editors can’t keep you from making a poor decision in your editing workflow (at least not yet). Replacing a sky, or relighting an image is only as convincing as the job you do. We spend our whole lives looking at light, we see the way it falls on a scene before us, and our brain instantly sends up signals when there is something off. If you choose a sky with the light coming from the wrong direction, your viewers will notice right away, even if they can’t articulate the problem in the image.

Learn to observe the way the light in your scene hits your subject, look at the temperature of the light, is it warm or cool? The more adept you become at seeing the light you are capturing, the better your images will look. Photography is the art, application, and practice of creating durable images by recording light, learning to see light can help every aspect of your photography.

Final thoughts

How do you feel about the impact AI editing programs have had on the photography industry? Do you use these programs in your editing workflow? Do you feel like they are cheating? I’d love to hear your thoughts in the comments.