Introduction

A good AI recolour image prompt should do more than say “colorize this photo.” When you want to recolour black and white photos, the prompt should explain what must stay unchanged and what kind of color result you want.

If your prompt is too simple, the AI photo colorizer may create colors that look too modern, too saturated, or historically inaccurate. If your prompt is structured clearly, you can get a more natural result for old family photos, portraits, landscapes, and vintage memories.

Basic AI Recolour Image Prompt Structure

A strong AI recolour image prompt should include five parts:

Original details: what should stay unchanged
Color style: natural, muted, vintage, realistic, or warm
People: skin tone, hair, eyes, clothing
Background: sky, buildings, trees, room, landscape
Texture: film grain, shadows, old photo feeling

Example:

Recolour this black and white photo into a natural realistic color photo. Keep the original face, identity, pose, clothing, background, lighting direction, photo texture, film grain, and vintage atmosphere unchanged. Use natural skin tones, soft clothing colors, realistic background colors, and balanced saturation.

Old Family Photo Prompt

For old family photos, the prompt should protect identity and emotional detail. Mention faces, age, pose, expression, clothing, and the original vintage texture.

Example:

Recolour this old family photo with natural, historically believable colors. Keep every person’s face, identity, age, expression, pose, hairstyle, and clothing shape unchanged. Add realistic skin tones, soft hair color, muted clothing colors, and gentle background tones. Preserve the original film grain, shadows, and vintage family photo atmosphere.

Black and White Portrait Prompt

For black and white portraits, focus on skin tone, facial detail, hair, eyes, clothing, and lighting. Avoid asking for overly bright color.

Example:

Recolour this black and white portrait into a realistic color photograph. Keep the original facial features, eye shape, expression, hairstyle, pose, and lighting unchanged. Add soft natural skin tone, realistic hair color, subtle eye color, muted clothing color, and a gentle vintage background. Do not over-saturate the image.

Vintage Landscape Prompt

For landscapes, guide the AI on sky, water, grass, mountains, buildings, and natural light.

Example:

Recolour this black and white landscape photo with natural realistic colors. Keep the original mountains, lake, trees, sky, reflections, lighting, and composition unchanged. Add soft blue-gray sky, natural green trees, realistic water reflections, earthy rocks, and subtle vintage film texture. Avoid modern HDR colors.

Follow-up Prompt

If the first recolour image AI result looks unnatural, use a follow-up prompt instead of starting over.

Example:

Make the colors softer and more historically realistic. Reduce saturation, keep the skin tones natural, make the clothing colors more muted, and preserve the original old photo texture.

Conclusion

The best AI recolour image prompts are clear, protective, and specific. To recolour old photos naturally, always describe what should remain unchanged and what kind of color style you want.

For better results, write prompts with identity, skin tone, clothing, background, lighting, texture, and saturation rules. This makes recolour photos online workflows more useful for family photos, black and white portraits, historical images, and vintage landscapes.

Reference

Tom’s Guide (2025) ‘How to colorize black and white photos with Google Gemini — no Photoshop required’. Available at: https://www.tomsguide.com/ai/google-gemini/how-to-colorize-black-and-white-photos-with-google-gemini-no-photoshop-required (Accessed: 3 July 2026).

Li, B. et al. (2026) ‘ColorFLUX: A Structure-Color Decoupling Framework for Old Photo Colorization’, arXiv. Available at: https://arxiv.org/abs/2603.28162 (Accessed: 3 July 2026).

Zhang, R., Isola, P. and Efros, A.A. (2016) ‘Colorful Image Colorization’, arXiv. Available at: https://arxiv.org/abs/1603.08511 (Accessed: 3 July 2026).