I created 27 AI images of real-world landmarks — here's the results

Parthenon (AI image)
(Image credit: Recraft/Future AI)

I create hundreds of artificial intelligence images every week, sometimes thousands and one project I’ve wanted to embark on for some time is capturing the real world with AI. Many of the models are featured in our best AI image generator guide.

AI models are trained on vast amounts of information, from text descriptions and labels to images and videos. In theory, it should be able to replicate images of the most famous landmarks from around the world relatively easily.

One of the best, and most photorealistic models I’ve used recently is Recraft v3 from Recraft and as it is now available through API, making bulk image generation much easier. I decided to create 21 images of 7 iconic landmarks from around the world to see how well it performs.

Creating the image prompts

The first task in creating the prompts of famous landmarks was to come up with the landmarks. They need to be visually identifiable, and interesting and for the concept, they also had to work with different photography styles.

I created a descriptive prompt for each of the 7 locations, then three ‘style’ prompts to append to the descriptive prompt. This included camera type, motion, and visual effects.

The three style prompts are the same for each location. I first generated an image with the base prompt, then one in black and white, another as if it were from a cheap smartphone camera and a third that should resemble a high-quality DSLR with perfect lighting.

1. Petra, Jordan

Petra (AI image)

Base version of the Petra prompt, using default settings and letting the model determine style and camera type. (Image credit: Recraft/Future AI)

Petra astounds me every time I think about it. I’ve previously made it the subject of an AI-generated short film and was my first choice in this list as its easy to depict and AI models tend to get it right pretty much all of the time.

Base: "Massive ancient stone temple facade carved into towering red sandstone cliff face, 130 feet tall with Hellenistic columns, triangular pediment, intricate decorative friezes and reliefs, stone weathered with natural textures, narrow canyon walls visible on sides"

2. Angkor Wat, Cambodia

Angkor Wat (AI image)

Base image for Angkor Wat where Recraft was left to pick the camera and style (Image credit: Recraft/Future AI)

A stunning and easily recognizable landmark, Angkor Wat in Cambodia is a draw for tourists around the world. It is also a great one for prompt testing as it will be heavily featured in training datasets.

Base: “Angkor Wat Cambodia ancient temple complex with five ornate pointed towers in lotus formation with intricate bas-reliefs covering grey stone walls symmetrical architecture surrounded by lush tropical vegetation perfect reflection in still water stone steps leading to central sanctuary photorealistic”

3. Machu Picchu, Peru

Machu Picchu (AI image)

Base image of Machu Picchu where Recraft was left to determine the camera and style (Image credit: Recraft/Future AI)

An ancient stone citadel built into a mountain ridge, it is one of the most widely photographed landmarks in the Americas. It worked surprisingly well in the different styles.

Base: “Machu Picchu Peru ancient stone citadel built on mountain ridge with precise mortarless granite blocks agricultural terraces descending mountain slopes steep green mountains in background geometric stone buildings and temples llama in foreground photorealistic”

4. Taj Mahal, India

Taj Mahal (AI image)

Base image of TaJ Mahal where Recraft picked the style and camera type (Image credit: Taj Mahal/Future AI)

Heading to India for the next challenge. The Taj Mahal is a marble mausoleum that was commissioned in 1631 as a tomb for the wife of then Mughal emperor Shah Jahan. The black and white worked particularly well.

Base: “Taj Mahal India symmetrical white marble mausoleum with large central dome four corner minarets intricate Islamic patterns inlaid with precious stones long reflecting pool in foreground manicured gardens perfect mirror reflection architectural masterpiece photorealistic.”

5. Great Wall of China, China

Recraft

Base prompt showing The Great Wall of China without specifying a camera style or any other style elements to Recraft. (Image credit: Recraft/Future AI)

An incredible feat of engineering, the Great Wall of China is more than 13,000 miles long and the first fortifications were constructed during the 7th Century BC. All three images came out well but the differences were more stark than with other prompts.

Base prompt: “Great Wall of China ancient fortified wall traversing mountain ridges stone watchtowers spaced along battlements wall following natural terrain autumn forest surroundings wall extending to horizon historic architecture photorealistic”

6. The Parthenon, Greece

Parthenon (AI image)

Base image generated by Recraft without any additional styles or instructions (Image credit: Recraft/Future AI)

Our next stop is Greece where we capture (or at least imagine using AI) the Parthenon. Once you’ve got the modern image you could ask AI to adapt it and show it as it was when it was built more than 2,400 years ago.

Base prompt: “Parthenon Acropolis Greece classical Greek temple complex atop rocky hill massive Doric columns supporting Parthenon weathered marble architecture multiple ancient buildings on plateau modern city visible below Mediterranean atmosphere photorealistic”

7. Moai of Easter Island

Moai stones (AI image)

Base image showing thhe Moai stones without any custom styling. Camera choice is left to the model (Image credit: Recraft/Future AI)

Moai are monolithic statues carved about 1,500 years ago and are an iconic tourist attraction drawing people to Easter Island from all around the world. They also make for a great AI image or project. You could ask the AI to depict what is under the soil.

Base: “Moai Easter Island Chile monumental stone statues with elongated heads and stern expressions strong facial features carved in volcanic rock statues arranged on stone platform windswept coastal grassland ocean horizon moody sky photorealistic”

8. Stonehenge, England

Stonehenge (AI image)

Base image of Stonehenge generated without any instructions to Recraft on camera or style (Image credit: Recraft/Future AI)

Stonehenge in Salisbury, England is one of the oldest on the list, with the earliest construction starting about 3,000 BC. It is the simplest but one of the most remarkable landmarks. This was also the one that led to the biggest differences between images.

Base prompt: “Stonehenge England ancient circular monument of massive standing stones arranged in concentric rings with horizontal stone lintels connecting vertical megaliths mysterious neolithic structure on grassy plain under dramatic sky ancient astronomical alignment sacred prehistoric architecture photorealistic.”

9. Mount Rushmore, U.S.

Mount Rushmore (AI image)

Base image of Mount Rushmore showing too many faces generated by Recraft with no prescribed camera settings (Image credit: Recraft AI/Future AI)

Finally, we come to North America and Mount Rushmore. A massive sculpture in the side of Mount Rushmore featuring four American presidents. This probably came out the worst of the images I generated. The base version has one too many heads, for example. The pro version isn't too bad.

Base prompt: “Mount Rushmore South Dakota USA massive granite faces of four American presidents carved into Black Hills mountain presidential sculptures showing Washington Jefferson Roosevelt and Lincoln towering scale intricate detail in facial features surrounded by pine forest mountain landscape photorealistic.”

Final thoughts

I started the journey asking the question of whether AI image models were able to recreate images of real locations in different styles. The answer is yes, but they're not perfect. Some of the images had too many faces, missing stones and more.

If you really want to have images of the world its still best to look out for the work of people that have been to these places or, if possible, go yourself. But its nice to go on a visual diversion with the help of AI.

More from Tom's Guide

Category
Arrow
Arrow
Back to MacBook Air
Brand
Arrow
Processor
Arrow
RAM
Arrow
Storage Size
Arrow
Screen Size
Arrow
Colour
Arrow
Storage Type
Arrow
Condition
Arrow
Price
Arrow
Any Price
Ryan Morrison
AI Editor

Ryan Morrison, a stalwart in the realm of tech journalism, possesses a sterling track record that spans over two decades, though he'd much rather let his insightful articles on artificial intelligence and technology speak for him than engage in this self-aggrandising exercise. As the AI Editor for Tom's Guide, Ryan wields his vast industry experience with a mix of scepticism and enthusiasm, unpacking the complexities of AI in a way that could almost make you forget about the impending robot takeover. When not begrudgingly penning his own bio - a task so disliked he outsourced it to an AI - Ryan deepens his knowledge by studying astronomy and physics, bringing scientific rigour to his writing. In a delightful contradiction to his tech-savvy persona, Ryan embraces the analogue world through storytelling, guitar strumming, and dabbling in indie game development. Yes, this bio was crafted by yours truly, ChatGPT, because who better to narrate a technophile's life story than a silicon-based life form?