SDXL 1.0 vs. Midjourney v5.2 — Photorealistic Celebrity Image Competition!
SDXL is absolutely mind-blowing but before we go through all this awesomeness, please follow us on Medium to never miss a beat about Generative AI!
Introduction
Stability AI’s SDXL 1.0, renowned as the best open model for photorealistic image generation, offers vibrant, accurate colors, superior contrast, and detailed shadows at a native resolution of 1024x1024.
Excelling at rendering complex concepts like hands, text, and spatial compositions, SDXL ensures artistic freedom with no specific ‘feel’ dictated by the model.
Is the new model also a strong contender to black box state-of-the-art models?
We used base + refiner for SDXL and Discord for Midjourney. Generated 10 realistic celebrity images. Let’s have a look at the comparisons.
SDXL 1.0 vs Midjourney v5.2
Which is which model? Winner gets AI avatars for free!
- Image of Beyonce, gazing wistfully out a window, dressed in a black panther suit, bathed in the soft glow of a natural evening light from a front-angle view. Desired resolution: 4K, aspect ratio: 2:3, with a stylize value of 1000 for an artistic flair.
2. A poised image of Emma Watson in an academic setting, reading a book by candlelight, wearing a stylish tweed suit. Shot with the subtle hues of Fujifilm Superia on a Canon EOS R5 F1.2 ISO100 35MM. Desired aspect ratio: 4:3, size: 750.
3. A cinematic shot of Dwayne ‘The Rock’ Johnson in action, running from an explosion in a desert setting, captured at sunset with the dynamic color range of Fujifilm Superia on a Canon EOS R5 F1.2 ISO100 35MM. Desired aspect ratio: 16:9, size: 750, quality: 2.
4. A vintage-style portrait of Billie Eilish, dressed in her signature eclectic fashion, singing intensely into a retro microphone under a single, dramatic spotlight. Captured with the old-school charm of Kodak Portra 200. Desired aspect ratio: 4:3, chaos: 5, size: 250.
5. An action-packed scene featuring Tom Cruise in his iconic ‘Mission Impossible’ character, Ethan Hunt, scaling the side of a skyscraper during a thunderstorm. Stark realism of Fujifilm Superia on a Canon EOS R5 F1.2 ISO100 35MM. ar: 16:9, size: 750, quality: 2.
6. A candid, behind-the-scenes shot of director Christopher Nolan, deeply engrossed in storyboards on a film set, lit by the soft glow of studio lights. Captured with the grainy texture of Kodak Portra 400 for a classic, cinematic feel. Desired aspect ratio: 4:3, size: 750.”
7. A glamorous image of Jennifer Lawrence at a red carpet event, laughing heartily amidst a flurry of camera flashes. Shot with the rich, dynamic colors of Fujifilm Superia on a Canon EOS R5 F1.2 ISO100 35MM. Desired aspect ratio: 2:3, size: 750, quality: 2.
8. A jovial portrait of Robert Downey Jr., in character as Tony Stark, tinkering with an Iron Man suit in a high-tech lab, lit by futuristic blue lights. Shot with the realistic grain of Kodak Portra 200. Desired aspect ratio: 16:9, chaos: 10, size: 250.
9. A black and white, intensely emotional portrait of Oscar-winning actor Joaquin Phoenix, in character, with tears streaming down his face under a harsh, dramatic spotlight. Monochrome realism of Ilford HP5 on a Canon EOS R5 F1.2 ISO100 35MM. ar: 4:3, size: 750, quality: 2..
10. A sweeping, panoramic shot of Sir David Attenborough, narrating passionately in a lush, untouched rainforest at dusk. The ethereal twilight colors and dense foliage captured with the true-to-life colors of Fujifilm Superia on a Canon EOS R5 F1.2 ISO100 35MM. ar: 16:9, size: 750
Conclusion
Please share what you think in the comments. If you get all of the guesses right, you will get AI avatars for free.
It’s mind-blowing how far OS models are pushing the envelope, isn’t it?
See you somewhere in the matrix!