Restore & colorize old photos — with decade-accurate color and a 3-pass workflow

Paste an old or damaged photo into an image-capable AI (ChatGPT, Gemini, Nano Banana, or Flux). Use a 3-pass approach — each pass does ONE job, which preserves more detail than asking for everything at once.

PASS 1 — RESTORE (repair damage without changing content):
'Restore this photograph. Repair scratches, tears, creases, folds, spots, and water damage. Reduce noise and film grain artifacts. Sharpen details and upscale to high resolution. CRITICAL: preserve every face, expression, and body proportion EXACTLY as-is — do not alter, beautify, or modernize any facial features. Do not change the composition, crop, or add any elements. Return the restored black-and-white image.'

PASS 2 — COLORIZE (add historically plausible color):
Take the restored result from Pass 1 and run:
'Colorize this restored photograph with historically accurate, natural colors. [SELECT YOUR DECADE PALETTE BELOW]. Apply color to skin (with natural variation — cheeks slightly warmer, under-eyes slightly cooler), hair, clothing, and environment. Keep the colorization subtle and photographic — it should look like a color photo from the era, not a hand-tinted postcard. Preserve all facial features and details exactly.'

DECADE-SPECIFIC COLOR PALETTES (add the right one to Pass 2):
• 1920s-1930s: Muted sepia-adjacent tones, desaturated earth colors, low contrast, limited bright color — clothing in dark navys, browns, greys, cream
• 1940s: Muted earth tones, military olive and khaki, desaturated reds and blues, warm skin tones, Autochrome-like softness
• 1950s: Kodachrome warmth — rich warm reds, golden yellows, deep saturated blues, warm skin tones with a slight amber cast, higher contrast
• 1960s: Saturated primaries — vivid reds, true blues, Kelly greens, Ektachrome-like clarity and saturation, cooler skin tones
• 1970s: Amber, avocado green, rust orange, harvest gold, burnt sienna, warm and slightly faded, Kodak Gold tonality
• 1980s: Cooler overall, with electric blue, magenta, and teal accents, brighter and more contrasty, slight color shift toward blue/green

PASS 3 — ENHANCE (final refinement):
'Enhance this colorized photograph. Gently sharpen facial details — especially eyes, eyebrows, and hair texture. Improve the tonal range (deepen blacks slightly, lift midtones). Ensure the colorization looks natural and consistent across the entire image. Do not change any facial features, proportions, or expressions. The final result should look like a well-preserved color photograph from [DECADE].'

FACE PRESERVATION — THE CRITICAL TECHNIQUE:
Faces are where AI most wants to 'help' by beautifying, symmetrizing, or modernizing features. Fight this:
• Include 'EXACTLY as-is' and 'do not alter facial features' in every pass
• If a face shifts: re-run that pass with 'Match the face in the original EXACTLY — same nose shape, same eye spacing, same jawline, same asymmetries'
• For group photos: point out each person — 'the woman on the left, the man in the center' — and say 'maintain each person's distinct facial features'
• Compare the nose, chin, and eye spacing between passes — these are where drift happens most

Tips: the 3-pass approach preserves dramatically more detail than 'restore and colorize this'; if the photo has a known location or context (e.g., 'this is a 1952 family photo in suburban Ohio'), include that — it helps the AI choose accurate colors for architecture, vegetation, and clothing styles; for severely damaged photos, do Pass 1 twice (once for structural damage, once for noise/grain) before moving to colorization.