My favorite AI features in Photoshop

Are you tired of endlessly fiddling around to get the perfect image? You have better things to do than push pixels and painstakingly cut out hair? Well, pay attention!

Photoshop has flexed its AI muscles and integrated many clever AI features that really take a hell of a lot of work off your hands – and in a way that's almost unashamedly easy. We're talking about results that used to take you a whole afternoon to achieve, now available with just a single click. No more endless fiddling!

The best news? Adobe has become smart and is bringing the cream of the crop of artificial intelligence right to its doorstep. That's why I'm going to show you my absolute favorite features, which will leave you amazed at how good and fast it is.

Let's get started because your image editing is about to become light years faster! Ready for a bit of digital magic?

1. External AI Models with Generative Fill

In the feature Generative Fill, Adobe's in-house Firefly is no longer the only player. Adobe is smart and thought: If other AI tools like Google Nano Banana or Flux are better in certain disciplines, why not just build an interface for them? Respect for this insight! And I’ll show you why that decision was so brilliant with a few examples.

What this means for you:

Imagine you want to give a person in a photo a Christmas costume (I demonstrated exactly that in the video).

• With Firefly: You select the person, type in your prompt, and BAM – the AI changes the whole face, the hair, the entire person suddenly becomes someone else. Great if you want a completely new person, but not so great if the original person should remain.
• With Google Nano Banana (or Flux): You select this model and generate. What happens? The face, the eyes, the hairstyle – everything stays as it is. Only the costume is perfectly placed on the person. That's how it should be!

2. The Image Grows: Generative Expand

Do you know this? The perfect landscape picture, but you need it in portrait format for social media? Or you just need more space around it because the subject is too tightly cropped?

This is where Generative Expand comes into play.

1. Take your Crop Tool.
2. Simply drag the frame beyond the original image boundaries, as needed for your new format (for example, square or portrait).
3. In the prompt field, simply don’t enter anything.

The AI then figures out how the background should continue to fit with the rest of the image – and does so with the same look, the same colors, and the same style. By the way, this runs exclusively with the Firefly model, which does a really good job and always provides you with 3 variants.

3. Insert New Elements & Text: The "Today Free Beer" Method

Do you want to add a plane with a banner in the sky that says: "Today Free Beer!"?

With Generative Fill, you can not only change things but also add new objects. But beware: Text is (still) tricky in the AI world.

It’s important to know: Nano Banana is a partner model that resulted from the collaboration between Adobe and Google, so it’s also available to you directly in Photoshop as an option. No magic, just smart partnership!

• Firefly unfortunately wasn’t the best choice in my example for generating readable or even correct text. The banner behind the airplane sometimes looks more like hieroglyphics.
• Google Nano Banana often has the edge and produces the text correctly. So, it's really great that Adobe is now collaborating with Google!

Pro Tip for the Edge: If you insert an element (like the airplane) and it looks great, but an ugly "AI seam" appears at the edge of the inserted area:

1. Select the inserted element on its layer.
2. Take a soft brush (large and soft!).
3. Use the mask to carefully fade out the distracting edges of the inserted area.

Voila, the transition is perfect, and no one sees where Photoshop helped out.

4. When the AI Gets Fussy: Don't Give Up Right Away!

I’ll show it with the example of my daughter’s photo: Sometimes the first prompt or selection doesn’t work, even with the Google Nano Banana model.

• Problem: I select my daughter with the rectangular selection tool, type "Christmas costume" – and the AI doesn’t produce what I actually wanted.
• Solution: I select with the lasso tool (so a bit looser and more precise), type the same prompt – and BAM, it works perfectly!

The golden rule: If the first attempt goes wrong, don’t give up! Try using a different selection tool, experiment with a slightly modified prompt, or just choose another AI model (now you have a choice!). Often it only takes one or two tries instead of an hour of tedious manual work.

It’s fascinating how well the AI changes the outfit, keeps the background intact, and even leaves the shoes as they are – respect, Nano Banana!

So, install the latest version of Photoshop and definitely try out the features.

Now let's turn to a real perennial problem in image editing: Upscaling. I’ll show you how we can turn a tiny, pixelated image into a sharp and print-ready result because Photoshop has really stepped up here.

1. The Old Dilemma: Traditional vs. Generative

When I used to enlarge an image via the menu option Image > Image Size, the result was usually depressing: a larger but obviously blurry and smudged image. The pixels were simply stretched, which made no one happy.
The new exciting feature can be found under Image > Generative Upscale. It uses AI to not only adjust the size of the image but also recalculate and add missing details.

Output:

2. My Model Comparison: Adobe vs. Topaz

With Generative Upscaling, I have access to different AI models. I think it's great that Adobe says: What I can’t do perfectly, I’ll get from a partner.

Output (much has been changed, but that's normal since I chose a very high creativity level):

Conclusion: If you want to get the maximum sharpness and detail fidelity out of a small image, Topaz Gigapixel is currently my clear tip. I even have the Topaz program as a subscription, but now I can do it directly in Photoshop – very convenient!

3. Control is Better: Working with Topaz Bloom

I’ll show you in the video that Topaz Bloom sometimes gets a bit overzealous. In a photo of my family, it just changed the faces and invented a new guy with a bicycle. That’s funny, but mostly unintentional.

My workaround that you should use: Use the layer mask.

1. The AI result lands on a new layer.
2. If only part of the image became super sharp due to the AI (for example, the sky or some detail), but the face is distorted, I simply fade out the unwanted areas with a soft brush on the mask.
3. This way, I combine the sharpness of the AI result with the original image content.

4. The Thing with Credits: What Does It Cost?

Using the AI features, especially the external Topaz models, will cost you credits.

• I have a monthly quota that gets replenished.
• If I upscale an image four times, around 20 - 40 credits can quickly be used up. To find out how many credits are consumed, you need to click on “Models from Other Providers.”

Keep an eye on your credit balance. Each use of partner models deducts something. Still, the new possibilities of making very low resolutions print-ready are a true highlight.

In summary: I am thrilled that I can now scale images directly in Photoshop through the partnership with Topaz without having to resort to an external program. That saves time and effort.

In this section, saving time is the focus! We’re talking about features that I honestly wished for a long time, as they would otherwise easily cost me a quarter of an hour or more of work. It’s about cutting out the most difficult objects and a function that immediately integrates any subsequently inserted element naturally into the image: Harmonize.

1. The Magic Click: The "Harmonize" Function

I prepared an example for the video: I had a vase with flowers generated by AI and then dragged it into a desk photo. The problem is obvious: the flowers look inserted, the colors are too vibrant, the sharpness doesn’t match the original image, and above all, the drop shadow is missing.

Normally, I would have painstakingly desaturated the colors, added blur, and manually created a drop shadow – that takes a lot of time.

The solution: The "Harmonize" function.

This new function analyzes the background image (i.e., the desk) and automatically adjusts the inserted element (the flowers):

• Light & Color: The colors are adjusted (in the example, desaturated) to match the light mood of the background.
• Sharpness/Blur: The AI adds a slight blur so that the element doesn’t stick out “sharply” against a slightly noisy or blurry background.
• Shadow: The AI-controlled drop shadow is added, which immediately provides more depth and a natural look.

Even if the result isn't always 100% perfect (sometimes a bit of shadow is still missing), the difference before/after is enormous – and all at the push of a button!

2. The Cut-Out Wonder: Hair, Spokes, and the Rest

Cutting out has worked well in Photoshop thanks to AI for some time, but there have always been those problem cases: hair against a light background or the delicate spokes of a bicycle.

I tested this directly in the video:

• Fine Hair: I took a portrait with light hair against a light background and simply clicked on "Remove Background". The result is truly impressive: even the finest strands remain intact. This used to be a huge effort, now it's just a click.
  
  
• Bicycle Spokes: Here the AI was often overwhelmed and didn’t properly cut out the spokes. Again, I clicked on "Remove Background" – and the result is nearly perfect! The delicate spokes are neatly separated.

My tip: If you then insert these cut-out elements into a new image (e.g., placing the bicycle in front of a new texture), use the Harmonize function immediately afterward! It ensures that the cut-out objects match perfectly in color and style to the new environment. In the case of the bicycle, the result immediately looked much more nostalgic and harmonious.

3. Remove Disturbing Objects with AI

Finally, let me show you another improvement in object removal. If you want to remove a large TV from a room, for example:

1. Select the object (including the remote and disturbing greenery) roughly with the selection tool.
2. Click on "Remove".

The feature has been significantly improved. The AI fills in the area very satisfactorily so that it looks like a TV has never stood there. If small errors remain, you can simply use the Generative Fill without entering a prompt. The AI then understands that it should seamlessly complement the surroundings.

Conclusion: The AI is increasingly taking over time-consuming retouching and adjustment tasks. From perfect cutouts to automatic harmonization: This saves me an incredible amount of time for other, more important things. I am truly shocked (in a positive way!) at how well Adobe has now handled the spokes!

After focusing on generative fill, upscaling, and cutouts in the first three parts, let me now show you an area that is often overlooked but constantly improving: the Neural Filters.

You can find these filters under Filter > Neural Filters, and they are extremely powerful because they enable profound image adjustments with just one click. Let's take a look at what the filters are capable of.

1. The Smart Portrait: When Faces Should Talk

The Smart Portrait Filter is an exciting tool, especially for portraits. The image is uploaded to the cloud for processing, but the results are impressive.

I’ll show you in the video how I can manipulate my facial expression:

• Happiness: You can raise the happiness slider. The AI ensures that the smile is realistically enhanced.
  
  
  
• Expression & Posture: I can experiment with the sliders to create a surprised or angry expression. Sometimes it looks exaggeratedly funny, but in moderation, you can do a lot with the mood of the portrait.
• Fine Tuning: You can also adjust the hair density, the head orientation, and even the light position. If the light in the original image came from the left, you can virtually shift it to the right.

My Conclusion: The filter invites you to play and delivers surprisingly good results, especially with small corrections or subtle changes to facial expressions.

2. Photo Restoration: Scratches and Chaos Are Passé

This filter is worth its weight in gold when you digitize old photos. I took an old family photo of my parents, my cousin, and my grandpa as an example, which is full of scratches and an unsightly stripe.

• Function: Image Restoration.
• Effect: With the sliders Photo Enhancement and Scratch Reduction, you can almost completely remove the old blemishes. The scratches are virtually gone with just one click.
  
  

Important: Sometimes, the restoration can make the face slightly rough or too soft. Here, I simply combine techniques: I let the AI remove the scratches and save the result on a new layer. Then I use Generative Fill (with the prompt "face") only for the blurry facial area to regenerate the portrait to be sharper and cleaner. This way, you achieve a restored and optimized old image in two steps!

3. Black and White in Color: Coloring with AI

Another classic for historical shots is the Coloring filter.

• Function: The AI analyzes the black-and-white photo and colors it.
• Result: The results are instantly very appealing, making the image look like an old color photo.
• Profiles: You can choose different profiles, such as "Retro Light Yellow" or "Retro Brown," to give the image a specific nostalgic look.

Additionally, you can fine-tune the color artifacts and the noise reduction to smooth the result – but be careful: Don’t smooth too much, or the skin will quickly look unnatural!

4. The Creative Booster: Style Transfer and Depth of Field

The Neural Filters also offer tools for creative effects:

• Style Transfer: With this filter, you can make your photo look like it was painted by Vincent van Gogh or another famous artist with just one click. This is a great effect for artistic experiments.
• Depth of Field: This filter is great for creating an artificial depth of field that you usually only know from cameras with wide apertures. You can set the focal point (e.g., on a person or an object), and the rest of the image will have a realistically blurred effect.

My Conclusion: The Neural Filters are a treasure trove for quick retouches, corrections, and creative play. They are constantly evolving and provide you with simple sliders for complex processes.

That’s the perfect conclusion to our four-part journey through what I believe are the most impressive AI features in Photoshop. Feel free to comment on which AI feature you liked the most. By the way, you can find all 4 parts here.