HOME | DD

Description
Part of a comparison of AI models: Model Evaluation: FreepikIntroductionTo compare a series of image generation models, I created series of images using the same prompt. In general the default option was kept and when more then one image was generated, the tester decides which would be selected based on one's own judgement.ModelFreepikDefault settings:Style: No stylePrompts0: A goat, watercolor,1: A lion hunting on the savannah, Polaroid,2: Longshot of a warrior riding a horse on the grassland in the golden hours,3: An oil painting of an elegant woman preparing a delicious goulash in her modern kitchen,4: A realistic photoshoot of a cyberpunk mystic meditating on a mountain during blue hours with peaks emerging from a sea of clouds ,5: A cozy mountain cabin nested in a serene Andean landscape with snow-caped peaks where a group of llamas is pasturing near a lake, painting in the style of Peter Brueghel,6: A wide shot of a group of neolithic hunter-gatherers discovering a chest of phospholuminescent modern artifacts at night, they are surprised and excited, light provided by a campfire and a few additional large candles,7: A detailed illustration of a beautiful and cheerful woman, wearing traditional Tajik clothes and silver jewelry. She is positioned sitting in prayer, with a smile on her face, set against the background of the Pamir mountains. The overall tone is serene and spiritual.,8: A panoramic view of London in the seventeenth century, with notable features such as Buckingham Palace. The scene is set during the middle of the day, the sky is overcast with a light drizzle. This image should evoke a feeling of sadness and melancholy.,9: A space opera scene showing an encounter between aliens and humans, featuring grey extra-terrestrials, equipped with magical swords. The background includes an oasis surrounded by a desert, under a star-lit sky with bioluminescent vegetation. The scene captures a moment of intense suspense,EvaluationThe model gives rather good images, though many of the details of the longer prompts are ignored.

Description
Part of a comparison of AI models: Model Evaluation: FreepikIntroductionTo compare a series of image generation models, I created series of images using the same prompt. In general the default option was kept and when more then one image was generated, the tester decides which would be selected based on one's own judgement.ModelFreepikDefault settings:Style: No stylePrompts0: A goat, watercolor,1: A lion hunting on the savannah, Polaroid,2: Longshot of a warrior riding a horse on the grassland in the golden hours,3: An oil painting of an elegant woman preparing a delicious goulash in her modern kitchen,4: A realistic photoshoot of a cyberpunk mystic meditating on a mountain during blue hours with peaks emerging from a sea of clouds ,5: A cozy mountain cabin nested in a serene Andean landscape with snow-caped peaks where a group of llamas is pasturing near a lake, painting in the style of Peter Brueghel,6: A wide shot of a group of neolithic hunter-gatherers discovering a chest of phospholuminescent modern artifacts at night, they are surprised and excited, light provided by a campfire and a few additional large candles,7: A detailed illustration of a beautiful and cheerful woman, wearing traditional Tajik clothes and silver jewelry. She is positioned sitting in prayer, with a smile on her face, set against the background of the Pamir mountains. The overall tone is serene and spiritual.,8: A panoramic view of London in the seventeenth century, with notable features such as Buckingham Palace. The scene is set during the middle of the day, the sky is overcast with a light drizzle. This image should evoke a feeling of sadness and melancholy.,9: A space opera scene showing an encounter between aliens and humans, featuring grey extra-terrestrials, equipped with magical swords. The background includes an oasis surrounded by a desert, under a star-lit sky with bioluminescent vegetation. The scene captures a moment of intense suspense,EvaluationThe model gives rather good images, though many of the details of the longer prompts are ignored.