Skip to main content
Stories Restored

How to Restore Old Photos for Free with Gemini AI (DIY Guide)

Restore and colorize old family photos in about a minute using Google's Gemini. No editing skills required. Just seven steps.

Old photos fade, crack, and yellow. Restoring one used to mean Photoshop or a $50 pro job. Now Gemini can do it free in about a minute. Here's how.

All you need is a Google account and a digital copy of your photo. The whole process runs in your browser.

Before you start: have a clean digital copy ready

Restoration only works as well as the photo you give it. Use a flatbed scanner if you have one — or follow a few simple tips when photographing with your phone.

See the full prep guide →

Steps

Too many photos for one-by-one? Skip the work →

1

Go to gemini.google.com

Open your web browser and navigate to gemini.google.com. You'll land on the main Gemini chat interface. In the top-right corner you'll see a Sign in button — click that to continue.

Gemini homepage showing the Sign in button
2

Sign in with your Google account

Sign in with any existing Google account — Gmail, Google Drive, YouTube, it's all the same login. If you don't have one, click Create account — it's free and takes under a minute.

Google sign in page
3

Select the Thinking model

Once signed in, look at the bottom-right of the chat box — you'll see a model selector (it may say “Fast” by default). Click it and choose Thinking. This gives Gemini deeper reasoning ability, which produces significantly better photo restorations.

Gemini model selector showing the Thinking option
4

Select the Create image tool

Click the Tools icon (grid of squares) next to the chat box. From the dropdown, select Create image. This activates Gemini's image generation and editing capability — you'll see it appear as a tag next to the input box once selected.

Gemini Tools menu with Create image selected
5

Upload your photo

Click the + button on the left side of the chat box. Choose Upload files if the photo is saved on your computer, or Add from Drive if it's in Google Drive. Select one photo — Gemini works best with one image at a time.

Upload files menu in Gemini
6

Paste the prompt and send

Copy the prompt below and paste it into the chat box alongside your uploaded photo. This tells Gemini exactly what to do — straighten, restore, enhance, and colorize. Then hit the send arrow.

This image is a photograph of a physical printed photo. Extract only the original photo content: correct any perspective distortion or tilt so the photo is perfectly straight, crop out all borders, frames, table surfaces, fingers, and surrounding environment. Then enhance the extracted image — improve sharpness, contrast, and remove any glare or reflections from the camera flash. Return only the extracted and enhanced photo with no surrounding context. Preserve the composition, people, and scene exactly. Reconstruct missing details naturally. Apply modern color grading while respecting the era of the original photo and infer the color if the image is black and white.

Gemini chat showing the uploaded photo and prompt
7

Download the restored photo

Gemini analyzes the photo and returns the restored, colorized image in a few seconds. Click the image to expand it, then click the download icon in the top-right corner to save the full-resolution version. That's yours to keep, print, frame, or share.

Gemini showing the restored colorized photo with download button

Common problems and fixes

Faces look different from the original

This happens when the source scan is too low-resolution or a face is partly covered. Re-scan at 600 DPI or take a sharper phone photo with even lighting, then run the prompt again.

Colors look unnatural or too saturated

Add “keep colors muted and period-accurate” to the prompt. If that still looks off, ask Gemini to return a black-and-white version instead.

Output is cropped or distorted

Crop the photo borders out before uploading. Gemini sometimes mistakes a frame, hand, or table edge for part of the image.

The image comes back unchanged

Confirm the Create image tool is selected (you should see a tag on the input box). Without it, Gemini just describes the photo in text.

What Gemini can't restore

Gemini is a generative model. It fills missing detail with a best guess based on the surrounding pixels. That works well for scratches, dust, and faded color. It works less well when the original information is gone for good.

  • Faces it has no reference for. A torn portrait missing half a face gets reconstructed, but the rebuilt half is a guess. It will look like a plausible person, not necessarily the right one.
  • Tiny, unreadable text. Newspaper clippings and document scans with very small type usually come back with invented words.
  • Heavily compressed or pixelated sources. If the original is a 200px JPEG screengrab, no model can pull real detail out of it.
  • Original artistic intent. A deliberately sepia-toned or hand-tinted photo gets re-colorized like any other. Note the original style in the prompt if it matters.

Got more than one or two photos? Running through the steps above for a whole box adds up fast. We can handle that — see how the bulk service works →

Common Questions

What if my restored photo doesn't look right?

Gemini's output varies. If the result looks off, try re-uploading with a cleaner scan, or adjust the prompt slightly. Heavily damaged photos sometimes need more than one pass.

Can I restore multiple photos this way?

Yes, but it takes a few minutes per photo. If you have a box of them, our bulk service does it in one upload.

How is this different from the paid service?

The DIY route is free and works well for one or two photos. The paid service uses hand-tuned processing, handles large batches, and lets us catch problems a one-size-fits-all prompt misses.