There's something about black and white that keeps history at a distance. I used AI to colorize a series of old public domain photos -- just to see what would change. The results were a mix of surprising successes and some quirks I wasn't quite expecting.
Rustic Farm Photo
I started my experimenting with a monochrome rustic farm scene. The Image Colorizer tool by DeepAI works by analyzing grayscale tones and predicting realistic colors based on learned patterns from millions of images. After uploading my image, the AI processes it in seconds and applies colors to each element of the photo.
For each of the following images, the original photo is displayed on the left, and the newly colorized photo is shown on the right.
Overall, the colorization turned out surprisingly well. The AI did a great job with key elements like the barns, sky, and grass fields. The barn's wood, the muted greens of the grass, and the soft sky all looked fairly natural and weathered, giving the scene an authentic rural feel. It didn't produce a random splash of colors, it felt like a realistic portrayal of a day from the past.
However, there were a couple of oddities. First, the grass subtly shifted in color as it went from left to right. It wasn't a dealbreaker, but it did stand out as a minor flaw. Additionally, the image had an overall sepia-like filter, giving it a vintage, almost overdone look. It didn't quite feel as realistic as I hoped -- more like a nostalgic "old photo" effect than a true colorization. Still, it was a solid first attempt.
Man Wearing Bowler Hat
Next, I decided to try a closer image of a man in a bowler hat. I was curious to see how DeepAI would handle the intricate details of clothing and skin tones.
The overall result was decent. The man's face came to life, with realistic skin tones and subtle shading. His flowers turned out particularly well, showing off a clear distinction between the leaves and petals. His clothing, while a bit rough around the edges, was mostly believable, with dark tones that fit the time period.
There were a few glaring issues, though. For one, there was a bright orange spot near the left side of the image, which definitely didn't belong. The hat had a brownish hue while the suit was colored black, which I'm fairly certain wouldn't have been the case when the photo was originally taken. The sepia filter also made another appearance, giving the whole image an antique vibe that felt out of place in a more realistic colorization attempt.
Related Why I Prefer Editing My Photos With AI Tools Over Manual Techniques
Letting AI handle the heavy lifting in photo editing has been a game-changer for me.
Posts Vintage Cars
Vintage cars can be tough to colorize, especially because of all the small details -- the paint, the rust, the intricate bodywork. I uploaded a photo of two classic cars parked in front of a shack, each with its own set of unique details. My expectations were greatly exceeded.
To my surprise, the result came out looking quite realistic. The AI did a solid job with the cars themselves. The metal parts had a nice, rustic quality, and the colors were crisp. The background looked natural and matched the overall tone of the image. I wasn't expecting this one to come out so well, especially with the level of detail involved. The final result almost looked like it could have been the original photo.
That said, not everything was perfect. There were a few subtle color inconsistencies that didn't sit right. For example, both of the cars had hood colors that didn't match their roof colors -- this could've happened due to wear and tear, but it felt the AI made too much of a guess. Small details like these prevented the colorization from being flawless, though it still looked impressive overall.
Old Fire Station
For this experiment, I uploaded a photo of an old fire station. I was curious to see how DeepAI would handle the small-town feel.
Once again, the results were fairly good (aside from one major anomaly, which I'm sure you've noticed). The colors were believable for the most part. The brick building came through with a natural reddish-brown tone, while the textures held onto their old, rustic appeal. The trim and roofing on the fire station were done well, and the colors didn't feel out of place.
The main problem, as I've come to expect by now, was the sepia filter. It's as if DeepAI defaults to this look whenever it's uncertain, particularly when dealing with blurry or softer photos. The sepia gave it a warm, vintage look but also made it feel less like a true colorization and more like a filter over an old photo. While the colors themselves weren't bad, I couldn't help but wish they had a bit more vibrancy and realism.
Bonus: Elderly Man With a Cane
For a more unique experiment, I decided to colorize a photo I originally found in color -- an elderly man with a cane taking a stroll. I converted it to black-and-white first, then uploaded it to DeepAI to see how the colors compared to the original.
The following image shows the original photo on the left. The right side shows the newly colorized photo -- the result of making the original photo black and white, and then re-colorizing it using DeepAI's Image Colorizer.
The result was surprisingly close to the original. The people's skin tones, the texture of their clothing, and even the background elements looked quite realistic. In fact, I was impressed at how closely it matched the ambiance of the original photo, despite the AI working from a black-and-white version.
That said, there were some noticeable mistakes. The Coca-Cola tent, which was originally red, came out blue in the AI-generated version. This was a bit silly, but I'll give DeepAI a pass here since the logo wasn't overtly legible. Additionally, the man's cane shifted in color from silver to yellow halfway down, which stood out as a clear flaw.
The woman's orange jacket also took on a blue hue around her left sleeve -- another quirk that didn't quite align with reality. Despite these hiccups, the colorization was generally impressive.
Bonus: Butterfly on a Flower
Lastly, I tested out a more abstract image: a butterfly on a flower. Since butterflies come in all sorts of colors, I was curious to see how the AI would interpret this one.
Similar to the previous experiment, the following image shows the original photo on the left. The right side shows the newly colorized photo -- the result of making the original photo black and white, and then re-colorizing it using DeepAI's Image Colorizer.
The AI was free to be creative here, and it did a pretty good job. The butterfly's wings were a blend of colors that, while not matching the original species, looked natural. The flower's petals also took on a rich, vibrant color, which felt organic and fitting for the scene.
My only real criticism here is that the colors didn't match the original species of the butterfly or flower. Of course, this wasn't a disqualifier, considering the wide range of colors these organisms can come in. The AI filled in the gaps well with its interpretation, so this is more of a minor note than a major flaw.
Benefits, Drawbacks, and Recommendations
It's clear that while the technology offers impressive speed and simplicity, the results can be hit or miss. As a result, the final product can sometimes feel inconsistent, varying in quality depending on the AI's interpretation of the image.
AI colorization is fast, affordable, and easy to use. It's especially great for historical photos, giving them a fresh and dynamic look. Whether you're an enthusiast looking to breathe life into old family photos or a historian seeking to bring a new perspective to the past, this tool offers a lot of creative potential.
Of course, the technology still has its limitations. AI colorization struggles with certain image details and lacks the cultural or historical accuracy that a trained eye might bring. Sometimes, the results can be a little too off-the-mark, leading to unrealistic or harsh color choices. Additionally, certain types of photos -- especially blurry or soft ones -- often end up with an overdone sepia filter instead of a true-to-life colorization.
To get the best results, stick to high-resolution, clear images. The more detail the AI can work with, the better it will perform. Choose photos with recognizable objects (like faces or landmarks) to guide the AI in making color decisions. Lastly, avoid images with excessive texture or noise, as these can confuse the AI and lead to unrealistic outcomes.
Related I Used AI to Bring Old Black-and-White Family Photos to Life
Ever imagined your old black-and-white family photos in color? I did, and used these AI tools to bring them back to life.
Posts
Colorizing old photos with AI is a fun experiment, and while it's not perfect, the results can be pretty amazing. With a bit of patience and the right images, you're more than capable of breathing new life into some forgotten moments from the past.
