PHOTO EDITING WORKFLOW

Refresh an old photo while keeping its character

Use a restrained edit to reduce dust, fine scratches and faded contrast in a scanned photograph. The goal is a cleaner interpretation that respects the original image. AI reconstruction is not evidence of what an obscured historical detail actually looked like.

Illustrative black-and-white studio portrait of an older man — Concept / collection artwork · visual inspiration. This preview is not a verified output from the preset below. Your result will vary.

YOUR STARTING POINT

What you’ll need

Start with the clearest scan you have. Keep an untouched archival copy and upload a working JPG, PNG or WebP up to 4 MB. Avoid aggressively sharpening the source before uploading; that can make damaged edges harder to distinguish from real detail.

Initial model: Nano Banana Pro
Output setting: 2K · 2:3
Current-plan cost: 50 credits / image

Try this workflow

Opening the preset uses no credits. Review the prompt and quote before generating. A signed-in account, active plan and enough credits are required. See plans.

The starter prompt

Edit this brief for your own subject. It is a set of instructions to try, not a guarantee that every detail will be preserved.

Carefully restore this scanned photograph. Reduce surface dust, fine scratches, and discoloration; gently rebalance faded contrast. Preserve the people, facial features, age, expression, clothing, objects, framing, and period character. Retain monochrome if the original is black and white. Do not modernize the scene, invent missing people or text, or over-smooth texture.

Work through the idea

01
Begin with surface damage
Ask to reduce dust and light scratches before changing color or rebuilding missing areas. Retain monochrome if the original is black and white.
02
Protect the period details
Name clothing, facial features, age, expression and framing as things to preserve. Avoid instructions that modernize the scene.
03
Review at full size
Inspect faces, hands, lettering and damaged areas. Keep both versions and label the new version as AI-restored when its provenance matters.

Before you download or publish

No people, text or objects have been invented in damaged areas.
Faces retain recognizable features instead of becoming smooth generic portraits.
Texture, grain and monochrome character remain appropriate to the source.