# Old Photo Restoration Workflow

> Repair scratches, restore faces and enlarge a treasured old photo.

Canonical URL: https://allimageai.com/workflows/restore-old-photo/

Website: All Image AI

Language: English

Order matters when restoring a photo: repair damage before restoring faces, or the face model treats scratches as facial detail and sharpens them. This workflow fixes that order for you.

## Included in one run

- 20 credits total; all required steps and final exports included.
- Estimated time: 3–8 min
- Scale: 2x
- Face fidelity: 0.75 by default, biased to preservation
- Colourisation: Optional, inferred

## Inputs

- Old photo scan: A flat scan at 300 DPI or better gives the restoration models the most real detail to preserve.
- Known details (optional): Example: The coat was dark green and the house had pale yellow walls.
- Reference image required: Yes

## Editing steps

### 1. Upload a scan

Scanning beats photographing the print; 300 DPI or better.

### 2. Repair damage

Brush over scratches and missing areas.

[Object Remover](https://allimageai.com/ai-tools/object-remover.md)

### 3. Restore faces

Fidelity first, so features stay true.

[Old Photo Restoration](https://allimageai.com/ai-tools/old-photo-restoration.md)

[CodeFormer](https://allimageai.com/models/index.md)

### 4. Optional colourisation

Inferred colour for black-and-white originals.

[Nano Banana 2](https://allimageai.com/models/nano-banana-2.md)

### 5. Resize and export

A deterministic 2× resize produces the final printable file.

## Frequently asked questions

### Are the colours accurate?

No, they are plausible inference. State real colours in the prompt if you know them.

### It no longer looks like the person.

Raise Face Fidelity and rerun — under-repairing is the safer failure.