I have seen these pictures a million times before – Netaji Subhas Chandra Bose examining his line-up of women soldiers from the Rani of Jhansi Regiment. Black-flag demonstrations in 1928, with “Simon Go Back” written on a banner above them. These images have been reproduced in our textbooks a thousand times and shown in every Independence Day slide show since we can remember, so much so that these historical events have lost their sense of reality and turned into mere symbols. For this year’s Independence Day, I have tried my hand at bringing some colour into these classic photographs using AI. Here is how I did it and the things to look out for while trying it yourself.
Also read: Happy Independence Day 2026: 50+ WhatsApp wishes, quotes and greetings to share with your loved ones
Colourisation amplifies whatever is already in the photograph, so a blurry, low resolution jpeg will come out looking worse, not better. If you’re pulling images off the internet, look for archive or museum sources rather than a random social media repost, since those tend to be more heavily compressed.
This is the base prompt I used across all three images:
“Colourise this black and white photograph realistically. Preserve the original composition, facial features, expressions, and details exactly as they are, don’t alter or reimagine any part of the image. Add natural, historically accurate colours based on the time period, clothing fabric, skin tones, and setting shown in the photo. Keep the lighting, shadows, and grain and texture of the original photograph intact. Avoid oversaturated or artificial looking colours, aim for a muted, realistic film colour palette consistent with the era. Do not add or remove any people, objects, or background elements from the original image.”
This is the step people skip, and it’s the one that matters most. Generic colourisation prompts tend to default to whatever the model assumes a given decade looked like, which isn’t always accurate. Adding specific historical detail pushes the output much closer to reality.
For the photograph of Netaji Bose inspecting the Rani of Jhansi Regiment, I added: “The uniforms are Indian National Army khaki, a muted olive-brown, not bright green. Keep the uniform colour consistent across all soldiers in the frame.”
For the 1928 Simon Commission protest photograph, I added: “The flags and banner are black, hand-painted with white lettering reading Simon Go Back. Preserve the legibility of all text in the image.”
Once the tool generates a result, zoom in and compare it side by side with the original. Two things to look for specifically:
First, skin tones. AI image tools have a tendency to lighten and smooth skin tones during colourisation, which can noticeably alter how a historical figure is depicted. If this happens, regenerate with a stronger instruction to preserve the original tonal range, or manually correct it in an editing tool afterward.
Second, hallucinated detail. Check that no people, objects, or background elements have been added or removed, and that any text in the image, banners, signage, uniform insignia, hasn’t been altered or misread. This is more likely on crowd shots and photographs with fine printed or handwritten text.
If the colour on a specific element is off, a wrong uniform shade, an oversaturated sky, go back into the same conversation and ask the tool to adjust just that detail rather than starting over with a new prompt. This keeps everything else about the image, which the tool likely got right, untouched.
Colourising archival photographs isn’t just a fun weekend project once actual historical figures and events are involved. Even with a carefully written prompt, the AI is making interpretive choices, a shade of khaki here, a slightly warmer skin tone there, that a viewer might mistake for historical fact rather than an AI’s best guess. Treat the output as an artistic interpretation of the photograph, not a restoration, and it’s worth saying so wherever you share it. That distinction matters more this Independence Day than it might on any other day of the year.
Also read: Samsung may be developing its first Galaxy over-ear headphones to rival AirPods Max