Copy-ready restoration prompt
Restore this photograph conservatively. Repair only visible scratches, dust, stains, fading, creases, and mild blur. Preserve every person's identity, face shape, expression, pose, age, hairstyle, clothing, jewellery, body proportions, and position. Keep the original crop, background, lighting direction, photographic era, and natural film or paper texture. Do not apply beauty retouching, modern makeup, modern clothing, face replacement, dramatic sharpening, or new objects. Do not invent text or confident detail in unreadable areas; keep ambiguous regions restrained and consistent with the source. Return a natural restored version that still looks like the same photograph.
Why each instruction matters
- Repair only visible damage narrows the task from re-creation to correction
- Naming identity and geometry reduces unwanted face, pose, and crop changes
- Preserving the era discourages modern lighting, fashion, and camera aesthetics
- Keeping texture avoids plastic skin and an overprocessed result
- Calling out ambiguity acknowledges that missing information cannot be recovered exactly
Use the prompt in a controlled workflow
- Start with the highest-resolution uncropped scan available.
- Describe unusual damage after the main prompt, such as a diagonal tear across the left sleeve.
- Generate a restrained repair before asking for colourization or strong upscaling.
- Compare the source and result at matching size, concentrating on faces, hands, text, and repeated patterns.
- Save the approved result as a derivative and keep the original scan unchanged.
What a prompt cannot guarantee
Language guides a model but does not create evidence. If an eye, hand, medal, sign, or background object is fully missing, the result can only be inferred from context and training data.
A dedicated photo-restoration workflow can be more predictable than a general image generator because the task and controls are narrower. Even then, verify important details.
Prompt phrases to avoid
- Ultra-detailed, perfect face, flawless skin, or beauty portrait
- Make it look like it was shot on a modern professional camera
- 8K detail when the goal is identity fidelity rather than output size
- Fix everything without specifying what must remain unchanged
- Historically accurate colour without references for the people, place, and date
FAQs
What should an old photo restoration prompt include?
Describe the defects to repair, the details that must not change, the desired restraint, the original era and crop, and a requirement to avoid inventing unreadable content.
Can a prompt guarantee the same face?
No. A prompt can reduce unwanted changes but cannot recover information absent from the input. Use a specialist restoration workflow and verify the result.
Should I ask for 8K or ultra-detailed output?
Not by default. Resolution slogans can encourage invented texture. Ask for faithful restoration first, then upscale the approved result separately.
Sources
Preservation and technical guidance reviewed for this article.
- arXiv: Old Photo Restoration via Deep Latent Space Translation
- U.S. National Archives: Digitizing Family Papers and Photographs
Related
Ready to restore your photos?
Upload an image, choose your settings, and preview results with a before/after slider.