Nano Banana & Photoshop Integration Question

[Shay_Studio4]

I used Nano Banana to add a missing person into a group photo. We had four people in the original photo, and a fifth person who couldn’t attend was photographed separately. Nano Banana eventually created a good result, but it took many attempts and long delays due to free-tier limitations.

Since I already have a paid Photoshop subscription and saw that AI image models (including Nano Banana / Gemini) are now integrated, I wanted to try doing this directly in Photoshop.

However, I couldn’t figure out the correct workflow.

I’m not sure how to properly upload two images and ask Photoshop to combine them.
I tried placing them on separate layers, but Photoshop AI kept generating incorrect results or replacing people instead of simply adding the missing person.

I would really appreciate a clear explanation of the correct method to do this in Photoshop - preferably with screenshots or step-by-step instructions.

Is this feature supposed to work the same way as it does on the Nano Banana website (where you simply upload both images), or is the process different inside Photoshop?

I’m attaching the two original photos, the final result I managed to achieve on the Nano Banana website, and two additional less successful attempts.

Thank you in advance for any help!

[Trevor.Dennis]

I have had good results placing both original and new element on the same layer using Gemini 2.5, but just recently, not so much so.

To get this composite done at least, use Remove Background on the soccer strip person. 

Fix the sizing with Free Transform

And hit Harmonize in the Contextual Task Bar

The lighting and colour is wildly different, so fingers crossed this will work.

Not too bad.

I like to overlap new image elements with some of the original content, as it gives it depth, and ties things together.

So use Select Subject's 'Select People option. Select the man with the red jumper, and copy to a new layer.

Move the new layer above the soccer player, and use Harmonize

I took more care of the sizing this time, which has left the soccer player semi-legless.

So I placed a copy merged layer at the top of the layer stack, selected the missing legs, and used Generative Fill with  the Firefly 3.0 model.

No need for a prompt. Just hit Generate.

Even as little six months ago this would have been much more difficult, and the results possibly not so good.

[Trevor.Dennis]

It's still not a very exciting photograph, and since you asked about Ai you could use Gemini 3.0 to change the lighting.  I used this prompt:

'Improve lighting and color grading. Make the light come from the right, and produce a more vivid and exciting image.'

Which produced this result:

More interesting, but over saturated and a bit too much contrast.  Lets see if we can improve it with the Camera RAW filter.

I reduced Vibrance. Gave it some negative Clarity, and reduced Highlights a touch.  I also used Generative Upscale to double its size.  There is an issue with the upper body of the man in red being in front of the soccer player, but his legs are behind. Either a very awkward lean, or the Ai has stuffed up.  I'll let you work that out, but you only need select both of their legs from the waste down, and instruct Generative Fill to 'Place the man in red's left leg in front' or similar.  Show us how you got on.

Good luck.

[Shay_Studio4]

Thank you very much for the detailed answer, but this is not the result I was looking for.
I needed everyone to look at the hat. In the end, with the regular Nano, I achieved an excellent result. The question is why it is impossible to achieve this result in Photoshop, if it is the same Nano.