Skip to main content
Participating Frequently
September 29, 2026
Answered

Split a color photo to black and white layer and a color information layer

  • September 29, 2026
  • 6 replies
  • 36 views

First of all, I want to colorize old black and white pictures. I think that the neural filter doesn’t do what I am looking for.

So I came up with the following ‘solution’.

I ask Gemini to colorize the picture. That works really well. But Gemini alters the poses, eyes etc. a little bit. I don’t want that.

Than I was thinking of: is it possible to open the Gemini colorized photo in Photoshop, somehow extract the color to a layer, discard the Gemini picture, open the original photo and use the color layer just created op top of the original picture. No changes to faces, poses or other stuff.

But how I searched the net, no satisfying answer. Is there a way how to do this?

Gr. Christiaan 

    Correct answer Spas K.

    You can’t really do that with any native tools. You can use the gemini image as a reference and sample the colors to color your image in Photoshop utilizing various tools.

    The color layer will still have shapes. Using a blending mode between the two images would work if the images matched in terms of shapes / outlines, but then you’d already have the properly colored image anyway.

    6 replies

    Inspiring
    September 29, 2026

    What you need to do is frequency separation. This makes 2 layers, one contains the colour information and one is the texture information. 

    But this is going professional direction, so maybe you google and watch some tutorial videos how to do it.

    c.pfaffenbichler
    Community Expert
    Community Expert
    September 29, 2026

    Have you tried working in Lab? 

    D Fosse
    Community Expert
    Community Expert
    September 29, 2026

    It's straightforward to set one layer to Color blend mode and another layer to Luminance blend mode. That's basic Photohop 101.

     

    But when you say it alters poses and other image elements, that obviously won't work because they won't be aligned. Which again shows that you can't use AI to "enhance" images. You always get a new and different image on the other end, and that's just the nature of how generative AI works. It's not a bug, it's a feature.

    Participating Frequently
    September 29, 2026

    I have tried that a couple months ago. I got a tutorial somewhere. I could make a similar export as what you can do with only the color layer in the neural filter of colorizing a photo.
    I’ve noticed that the color information layer is not that detailed. And the alining worked quite well. But I can’t find that tutorial anymore.

    Spas K.
    Community Expert
    Spas K.Community ExpertCorrect answer
    Community Expert
    September 29, 2026

    You can’t really do that with any native tools. You can use the gemini image as a reference and sample the colors to color your image in Photoshop utilizing various tools.

    The color layer will still have shapes. Using a blending mode between the two images would work if the images matched in terms of shapes / outlines, but then you’d already have the properly colored image anyway.

    Participating Frequently
    September 29, 2026

    Thank you for your answer.