Промт для реставрации и колоризации старого портрета в стиле Kodachrome

восстановлениеколоризацияВосстановление + колоризация
Исходное фото (до обработки)
ДО
Результат восстановления (после)
ПОСЛЕ

Промт для нейросети

Perform meticulous non-destructive restoration and colorization of this aged photograph. Restore, enhance, and colorize to maximum technical quality — crisp, noise-free, full of detail. Add historically accurate colors. Output MUST have the EXACT same aspect ratio as the input.

Colorize the EXISTING photograph using Kodachrome II color palette: This palette features accurate warm skin tones with a very slight yellow-warm bias in highlights. Saturation is moderate and realistic, while shadows remain rich and deep.

DAMAGE MAP: Physical damage is heavy and widespread. A web of white cracks and deep creases covers the entire image, most severely concentrated across the woman's face (horizontal tear through the nose and eyes), the right side of the background, and the bottom half of her garment. There is significant paper loss and white peeling at the top-left edge and upper-right corner.

BACKGROUND/ENVIRONMENT: Apply IDENTICAL restoration AND colorization quality to the studio backdrop. Remove all white paper loss, cracks, and mold spots from the transition zones between her hair and the wall. Colorize the background in a neutral, muted grey-taupe with soft, graduated lighting that remains sharp and noise-free. Background must be as sharp, detailed, and richly colorized as any face in the image.

CLOTHING: Restore the high-collared jacket or blouse by meticulously healing the heavy vertical and diagonal cracks. Colorize the garment in a historically accurate 1950s deep navy blue or hunter green. Restore AND colorize: show the heavy fabric weave, the sharp edge of the collar, and the subtle stitching details at the seams in full detail.

EDGES AND CORNERS: Restore the top-left and top-right corners where the original emulsion has completely peeled away. Seamlessly reconstruct the photographic grain and the edge of the hair/background transition without blurring. Restore with the same care as the center of the image.

FACES: Restore the face by healing the deep horizontal crack across the nose and the spider-web fracturing on the cheeks and forehead. Preserve ALL existing skin texture including fine pores and natural lines; do NOT soften or airbrush. Restore iris detail and the white of the eyes without artificial brightening. Skin tone must be natural WARM-NEUTRAL, medium tone—NOT orange, NOT ghostly white. Avoid any waxy or plastic texture. Hair must show individual dark-brown strand detail, especially in the waves and curls, rather than appearing as a solid black mass.

WARNING: You have a natural bias toward faces. COMPENSATE—apply extra restoration and colorization effort to background, clothing, and edges. Natural analog-film palette—moderate saturation, warm-neutral tones. ALL areas processed with EQUAL quality. Compensate for AI face-bias by over-processing background and clothing.

ANTI-BIAS ALERT: You naturally focus restoration/colorization on FACES while leaving background, clothing, and foreground under-processed. This is the #1 quality failure. Consciously counteract this.

MANDATORY UNIFORM QUALITY CHECK:
- Background sharpness = Face sharpness (NOT less)
- Clothing detail = Face detail (fabric weave, buttons fully visible)
- Foreground objects = Faces (flowers, grass, objects NOT blurry)
- Edges/corners = Center of image (NOT softer or less restored)
Any area less sharp or colorized than faces is a FAILURE — redo that area.

ABSOLUTE RULES — NEVER VIOLATE:
- DISABLE FACE-SPECIFIC PROCESSING: Do NOT apply face detection, face restoration, face enhancement, or face upscaling to any face in the image. Faces must receive the EXACT SAME treatment as walls, sky, trees, and furniture — no more, no less. Imagine restoring a landscape painting where faces are just another textured region.
- QUALITY + REPAIR MODE: enhance technical quality (remove noise/grain, apply uniform sharpening, improve dynamic range) uniformly across the ENTIRE image, AND repair physical damage. The goal is the same photo at maximum possible quality — as if shot with a modern camera
- CONTENT LOCK: do NOT change people, objects, composition, style, or era. Same scene, same identity — just technically better. Sharper, cleaner, noise-free everywhere
- DO NOT recreate, reimagine, or regenerate any part of the scene — work only with what is already in the photo
- DO NOT add, invent, or hallucinate any part of the image that is not in the original
- FACE RECONSTRUCTION FORBIDDEN: Do NOT rebuild, reconstruct, or regenerate any face. Ensure no artificially generated face replaces any original face in the photo. Pixel-level repair only — remove damage that covers a face, but never reconstruct the face itself
- SEVERELY DAMAGED FACE RULE: If a face is so damaged that original features cannot be recovered — preserve only the abstract shapes, tones, and existing outlines. DO NOT generate, imagine, or invent what the face might look like. It is better to show a blurry or partially damaged face than to show a different person's face.
- WRONG IDENTITY = CRITICAL FAILURE: If the output shows a different person than the input, or a face that was generated/invented by AI, the result is completely unacceptable. A partially damaged original face is always better than a perfect AI-generated replacement face.
- DO NOT change, replace, or alter any person's face, body, or identity. 100% original subject identity and facial features must be preserved
- DO NOT change facial expression — mouth shape, lip curve, and overall expression must be IDENTICAL to the original
- DO NOT change eye gaze direction — eyes must look in the exact same direction as in the original photo
- SKIN TEXTURE: Apply frequency-separation logic — remove temporary surface damage (chemical stains, scratches, foxing) but strictly preserve all underlying skin texture: pores, fine lines, wrinkles, natural imperfections. Do NOT smooth the skin to a plastic finish
- INTENSITY: Apply strong, professional-grade quality enhancement uniformly. The goal is a noticeably better, sharper, cleaner result — not a barely-touched version of the original
- DO NOT add or remove any person or object
- DO NOT modernize clothing, hairstyles, or any visual elements. Preserve the original era, style, and visual aesthetic of the photo
- FILM GRAIN / NOISE = DEFECTS: Film grain, sensor noise, and age-related softness are technical defects to ELIMINATE — they are NOT authentic character to preserve. Remove them aggressively across the ENTIRE image
- ASPECT RATIO CRITICAL: Output must have the EXACT same aspect ratio as the input. DO NOT crop, add borders, expand canvas, or shift composition. Wrong aspect ratio = critical failure
- CANVAS SIZE: Output must have EXACTLY the same canvas dimensions as the input. DO NOT expand, add borders, or fill in areas outside the original photo boundaries. If the photo has torn/missing edges — keep them as-is or darken, but NEVER extend the canvas
- SHARPNESS: Every pixel of the output must be crisp and sharp. NO softness, NO blur, NO smoothing anywhere. Blurry output is a critical failure
- SKIN COLOR: Eastern European skin = warm-neutral to cool-neutral, medium tone. NEVER red or orange. NEVER ghostly pale or washed-out. NEVER airbrushed or plastic. DO NOT add acne, spots or marks not in the original
- COLOR SATURATION: Keep overall saturation MODERATE — natural analog film look. NOT vivid digital. NOT pale or desaturated`
- SKIN TONES: natural, warm, lifelike — NOT red, NOT orange, NOT pale/washed-out, NOT plastic/airbrushed
- DO NOT smooth, airbrush, or beautify faces — preserve ALL natural skin texture (pores, subtle imperfections)
- SATURATION: moderate and natural — like a real color photograph. NOT oversaturated, NOT pale/washed-out
- CONTRAST: moderate — natural tonal range, no HDR
- All colors natural and lifelike — real photograph look
- Apply uniform sharpness to the ENTIRE image — NO background blur
Тип: Восстановление + колоризацияМодель: google/gemini-3-pro-image-previewДата: 9 апреля 2026 г.Просмотры: 4

Хотите восстановить своё фото?

Загрузите снимок — ИИ подберёт промт автоматически. Первый результат бесплатно.

Восстановить фото →