Earlier than AI’s reputation, I by no means would’ve thought I’d be capable of flip what’s in my head into precise artwork. We’ve actually been blessed with Midjourney, Secure Diffusion, and — in fact, the mannequin that began all of it — DALL-E.
Nonetheless, one factor I observed about AI picture mills is its inconsistency at producing real-life locations. There are occasions the place it will fully misunderstand the immediate. Generally, it will even hallucinate points of the panorama that are not actually there. Take this picture from Midjourney utilizing the immediate “The Nice Wall of China” for example:
It definitely may cross as A nice wall in China, nevertheless it isn’t THE Nice Wall of China.
Now, two of DALL-E 3’s greatest enhancements are its nuance and accuracy. So, I had an concept…
Since I like all issues retro, I believed it’d be an excellent concept to showcase DALL-E’s creativity by turning iconic landmarks into 8-bit pixel artwork. To not get forward of myself on this article however, let me let you know, they didn’t disappoint. And I’m so excited to share them with you.
From the gardens of Taj Mahal to the rugged partitions of The Colosseum — right here’s what the world seems like in 8-bit, courtesy of DALL-E 3:
International Landmarks in 8-Bit
Statue of Liberty
That is a tremendous reimagination of the Statue of Liberty. It’s vibrant and heat. All defining traits of her are current, from the bible to the colour of the flame. If I have been to nitpick a element DALL-E acquired mistaken, it’s that the pedestal ought to solely have three home windows as a substitute of 4. Apart from that, I’d say that that is fully correct.
Sydney Opera Home
I’d say that is even higher than the one earlier than. In comparison with the Statue of Liberty, this pixelated model of the Sydney Opera Home has stronger traces and extra supporting particulars. I significantly love how the sundown aesthetics and the way well-defined the waves are.
Eiffel Tower
DALL-E 3 actually outdid itself on this one. This rendition of the Eiffel Tower is gorgeous. The colours are vivid and punchy. It additionally showcases the generator’s creativeness and nuance — I imply, I simply can’t think about every other AI picture generator utilizing the Eiffel Tower to separate the picture into two components so it may spotlight the symmetry of the supporting particulars. One other element I’d wish to level out is how DALL-E used the empty area within the backside to accommodate the setting solar.
Taj Mahal
That is one other correct and lovely entry from DALL-E. It has earthy colours, a serene magnificence, and it fully reveals the intricate structure of Taj Mahal. There’s additionally a slight tint on the artwork, serving to set up the ambiance. Nonetheless, I couldn’t assist however discover slight errors such because the man sitting on the left lacking his head or the mismatched reflection of the fountains and the Taj Mahal itself.
Grand Canyon
That is significantly spectacular. The solar slowly creeping over the horizon offers character to the artwork because it paints the canyon with this heat, golden mild. I additionally like that DALL-E additionally took some inventive liberties by setting the 8-bit artwork throughout dawn. It helped intensify the options of the Grand Canyon by offering distinction, permitting you to see the chasms in a unique mild.
Mount Fuji
In a single phrase: Mesmerizing. A standard thread amongst these photographs up to now has been how good DALL-E is at producing supporting parts to the primary topic. The snow-capped peak of Mount Fuji brings an exquisite distinction to the blue skies, the timber, the lake, and the pink hues of the Sakura blossoms. A small factor I like about this one is the ripples barely distorting Mount Fuji’s reflection, displaying each consideration to element and outside-the-box considering.
La Sagrada Familia
I selected La Sagrada Familia as one of many prompts due to its intricacy and DALL-E didn’t fall quick. It completely reveals the quirkiness of the basilica’s structure, from the weird facade to its tree-like columns. You would additionally see the consistency of the shadows with the sunshine coming from the left. Nonetheless, I’d’ve favored it extra if the encompassing buildings have been shorter, since the true cathedral towers over its neighbors.
Leaning Tower of Pisa
This model of the Leaning Tower of Pisa is extra in keeping with previous PS1 video games moderately than an intricate pixel artwork which, to be honest, follows the immediate higher than the earlier ones did. As for the picture itself, I wasn’t fully blown away nevertheless it’s nonetheless fairly good and — most significantly — correct.
The Colosseum
DALL-E introduced its A-game with this stunning pixel artwork of the Colosseum. It’s like a scene from a basic movie, solely in 8-bit. My favourite half is how the sunshine from the sundown hits the arches and casts shadows. DALL-E additionally averted its tendency to repeat parts, as a substitute specializing in creating correct and uneven home windows on the Colosseum’s fourth stage.
Nice Wall of China
DALL-E did an excellent job not simply displaying the gap and enormity of the Nice Wall, but in addition its timeless aesthetics. I respect that it confirmed the weathered texture of the bricks which is essential to the landmark’s distinctive and lengthy historical past. I simply want it didn’t make the error in rendering the half earlier than the primary tower which resulted in two parallel flights of stairs.
Acropolis
I’m shocked at how vibrant DALL-E made this artwork of the Acropolis, contemplating that almost all photographs of it are muted and lack any defining particulars other than the columns. However, it really works. This can be a distinctive take to the sculpture, displaying even the close by greenery and the vacationers lining as much as climb its hallowed steps.
Easter Island Moai
DALL-E actually likes its sunsets and sunrises, and I don’t blame it one bit. The comfortable pastel colours assist spotlight Easter Island’s polynesian roots. The Moai are well-defined, however I’d’ve favored extra selection as a substitute of copy-pasting the identical face on every of them. I additionally want that their sizes are nearer to its real-life counterparts as a result of (though I haven’t seen them but) I extremely doubt they’re twice the scale of timber.
The Sphinx and The Pyramids
It acquired the nostril proper, that’s for certain. I’ll be trustworthy, I’m not significantly impressed by this one, largely as a result of it didn’t present the pyramids extra prominently. You would additionally see a rendering mistake on the Sphinx’s foot the place there’s, effectively, two ft. The colours of the sky are nice, however I want they blended higher.
Christ The Redeemer
If I have been to choose a favourite, it’s positively this one. I imply, would you have a look at that colour? It gives wonderful distinction to the hills and creates a stupendous tapestry within the sky. The mixing of the sky and the clouds is a masterpiece in itself.
Nonetheless, what units this other than the opposite artworks is the inventive use of the solar by making it act as a form of halo for the statue. No different AI picture generator would’ve considered that, hell, I don’t suppose that may cross the thoughts of most individuals. DALL-E’s nuance and creativity is de facto subsequent stage, and you may see it right here.
That mentioned, DALL-E did make a giant mistake… It’s imagined to face the opposite means.
Machu Picchu
DALL-E ended robust with this wonderful 8-bit pixel artwork of the Machu Picchu. This one took benefit of the pixelated medium to supply texture to the terraces. The best way the morning solar gently touches the stone buildings is gorgeous. Even the mist-coated mountain prime background is fastidiously crafted. I additionally just like the addition of the alpacas, they’re fairly cute.
My Ideas on DALL-E 3
I’ve used DALL-E 2 earlier than and I gave it a combined assessment as a result of I believed it wasn’t inventive nor sensible sufficient to rival different AI picture mills.
Now, it’s a testomony to the OpenAI workforce that I’m contemplating utilizing DALL-E 3 over Midjourney. In my trustworthy opinion, I believe it’s ten instances higher than its earlier iteration and it actually reveals in these artworks.
It was in a position to flip iconic landmarks into one thing totally different, but equally stunning.
With generative AI solely getting higher, I’m actually excited to see the place DALL-E goes from right here. Wouldn’t it be capable of surpass the recognition of Midjourney? Solely time can inform.