◆ Photoshop · generate

Colourisation Memory Mismatch

We've colourised three archival photos this year and each time a client pushed back saying it felt made-up. Diagnose w…

1height
1tasks
1roles do it

The same job, four prompts

do it · improve it · decide · become
DBecome — “help me grow”We've colourised three archival photos this year and each time a client pushed back saying it…+
We've colourised three archival photos this year and each time a client pushed back saying it felt made-up. Diagnose where our colourisation process loses authenticity, and recommend one research habit and one restraint rule I should adopt so future colour references honour the original photo and still support the campaign's palette.
when the reply comes backPush once: ask it to sharpen the weakest part, and to say what it assumed.

How to do it

the tool · the steps · what to avoid
The tool
Generative Fill

Generative Fill uses AI to add, expand, or remove content from your images based on text prompts.

  1. 1

    Open your image by selecting File > Open.

  2. 2

    Select the area you want to transform using a selection tool.

    Best practiceFor expanding the canvas, use the Crop Tool to extend the image boundaries first.

  3. 3

    Click the Generative Fill button in the Contextual Task Bar.

    NoteYou can also access it from Edit > Generative Fill, or Select > All then Edit > Generative Fill.

  4. 4

    Type a prompt describing the object or scene you want to generate in the text-entry field.

    NoteYou can leave the prompt field empty to have Photoshop generate content based on the surrounding image.

  5. 5

    Click Generate.

  6. 6

    Review the generated variations in the Properties panel.

  7. 7

    Select your preferred variation from the options.

  8. 8

    Save your image by selecting File > Save As or File > Save a Copy.

Generative Fill creates new content on a separate generative layer, allowing non-destructive edits.

Generative Fill functionality is integrated within Photoshop and is not directly exposed as a standalone API for external automation.

Glossary

words on this page
Contextual Task BarA floating bar that provides relevant tools and actions based on the current selection or task.ExampleWhen you select an object, the Contextual Task Bar appears with options to duplicate or delete it.
Generative FillAn AI-powered feature that creates new image content based on text prompts or existing selections.ExampleUsing Generative Fill, you type 'add a sandy beach' to extend the ocean scene in your image.
AdjustmentsNon-destructive edits applied to an image to change its color, tone, or other visual characteristics.ExampleYou can use Adjustments to brighten a dark photo or change its color balance.
Save a CopyTo save the current document as a new file, leaving the original document open and unchanged.ExampleYou can 'Save a Copy' of your edited photo to share, while keeping the original file untouched.
TransformTo change the size, shape, or orientation of an image layer or selection.ExampleYou can transform an image layer to rotate it clockwise or make it much smaller.

The real tasks

1 of them
We keep tempting colourisation because it's pretty, then the client says it doesn't feel 'true' to memory.

Who does this

1 roles
Art Director

Questions people actually ask

with the jobs and tasks they touch

Sometimes, the AI-generated pixels don't have enough detail or proper blending, causing print quality issues at large scales.

Always review your extension at 100% zoom to see if the AI blended pixels smoothly. Fix any rough edges or mismatched areas before printing.

Use the 'Match Evening Mood' or 'Replace Sky' tools to better blend the extension with the original image. Adding subtle color adjustments helps too.

Relying on the AI to perfectly blend without manually checking or refining the extension can lead to visible seams or mismatched areas in large prints.

A photo editor or retoucher doing large-scale prints will notice these failures during final proofing or print setup.

AI can create a convincing extension, but for large prints, manual touch-up or blending is often needed to ensure quality at final size.

Yes, it's related because both involve creating or blending new pixels, but generated extensions focus on AI inventing missing parts for printing.