Сравнение Stable Diffusion 3.5 Large, Midjourney 6.1, Flux 1.1 Pro

Всем привет! В своей деятельности я иногда использую нейросети, создающие изображения. Они постоянно обновляются, поэтому такие сравнения быстро устаревают. Не так давно Midjourney получил версию 6.1, в начале месяца Flux Pro обновился до 1.1, и вот на днях вышел Stable Diffusion 3.5.

В этой статье я хочу сравнить их, используя 10 разнообразных промптов, чтобы и для себя, и для вас прояснить их различия и определить, кто из них лучше. Для генерации изображений в этих нейросетях я буду использовать телеграм-бота @bratuha2bot которым периодически пользуюсь.

Сразу хочу заметить, Midjourney всегда выдает 4 изображения, поэтому я решила не добавлять все 4, а выбрать 1 самое интересное на мой взгляд.

Итак, давайте начнем!

Промпт: A serene landscape at sunrise, featuring a calm lake reflecting the vibrant colors of the dawn, surrounded by majestic mountains that rise dramatically against the colorful sky, with gentle mist rising softly off the water, and a lone swan gliding gracefully across the surface, creating a sense of tranquility and peace in nature. 16:9

Stable Diffusion 3.5 Large

3d30ffa6d8d49980e1b0ec5fbeca99c5.jpeg

Midjourney 6.1

f790e2d3f18266e6848e6459d83e0b06.png

Flux 1.1 Pro

bb22461069f7ac83ff2f74c3ccb62c7c.jpeg

Промпт: A highly realistic portrait of a young woman with long, flowing hair cascading down her shoulders, captured in soft natural lighting that highlights her delicate features, her expressive eyes reflecting a world of emotion and depth, adorned with minimal and tasteful makeup, and wearing a delicate floral dress that complements her radiant smile, creating an inviting and warm atmosphere. 9:16

Stable Diffusion 3.5 Large

f778e25bc2246cba521bc9a04368a048.jpeg

Midjourney 6.1

bc12453174bf07f4ffde578e6b98c8bb.png

Flux 1.1 Pro

f584633cc22b9a88cd2e899829a33785.jpeg

Промпт: A futuristic cityscape at night, bustling with flying cars navigating through the air and neon lights illuminating the streets below, showcasing towering skyscrapers adorned with dynamic holographic advertisements, with a vibrant crowd of diverse people walking along the illuminated sidewalks, and the phrase 'Добро пожаловать в будущее' glowing above, embodying the energy and excitement of modern life. 16:9

Stable Diffusion 3.5 Large

5b109d5a492e5ff8cb6c53fc314b844f.jpeg

Midjourney 6.1

d4f7c1b7140d9a3889ba4720bf20386b.png

Flux 1.1 Pro

07a68456d44cd4be4054be65e887ea8c.jpeg

Промпт: An elegant book cover design featuring a mystical forest landscape at twilight, with ancient trees towering majestically and vibrant foliage framing the scene, surrounding a shimmering lake that reflects the last light of the day, and an ethereal figure standing on the shore, gazing up at the twinkling stars above, with the title 'Whispers of the Forest' elegantly scripted at the top in a whimsical font that draws the reader into a world of magic and adventure. 3:4

Stable Diffusion 3.5 Large

5fdf1b1db58ae2e3df43da7a9cbaa6b4.jpeg

Midjourney 6.1

64d79052e986261f4bbcd7c5467f0f34.png

Flux 1.1 Pro

27402b0d3ec3656415a8b8c0e48623bf.jpeg

Промпт: A breathtakingly realistic image of a grand historical building, its classical architecture standing proudly and illuminated by the soft golden hour light, showcasing intricate details in the stonework and elegant columns, surrounded by a lush garden bursting with colorful flowers and greenery, with a beautifully crafted sign reading 'Наследие прошлого' hanging at the entrance, emphasizing the building’s rich history and cultural significance. 16:9

Stable Diffusion 3.5 Large

27c05f1e07a584ad6bdebc176383a8e7.jpeg

Midjourney 6.1

fd66a9a9a6bef8b857f51087579557d8.png

Flux 1.1 Pro

08cfcfdb80f3a5bb67a3d86509255171.jpeg

Промпт: A vintage-style poster of a lively retro diner from the 1950s, featuring a bright red and white color scheme that evokes nostalgia, with classic chrome-plated cars parked outside and cheerful waitstaff serving delicious milkshakes and burgers, all captured in vibrant colors and dynamic poses, with the slogan 'Good Vibes & Great Food!' prominently displayed in bold typography, inviting everyone to enjoy a taste of the past. 16:9

Stable Diffusion 3.5 Large

739763527e5da83c9bf7739b1fe4ce13.jpeg

Midjourney 6.1

312136167120a508a4b1bbad02001090.png

Flux 1.1 Pro

f836fae721f2a0eb92f6dc5a84887c73.jpeg

Промпт: A vibrant underwater scene depicting a colorful coral reef teeming with a diverse array of fish, as sunlight filters through the crystal clear water, illuminating the various marine life and creating a dance of light and shadows, with a diver gracefully exploring the depths, capturing the beauty and complexity of the ocean environment, showcasing the delicate balance of life beneath the waves. 16:9

Stable Diffusion 3.5 Large

03a5512d8236c1ec570e4c31eff94968.jpeg

Midjourney 6.1

fe757d1648f9b45cdc0407348e5cbbc1.png

Flux 1.1 Pro

0d3ef6267fb1a174088fa364b420908f.jpeg

Промпт: A surreal dreamscape where a giant clock tower melts slowly into a shimmering river of stars, with whimsical floating islands covered in lush greenery and colorful flowers, and fantastical creatures like flying elephants and oversized butterflies fluttering gracefully around, creating a sense of wonder and whimsy that invites the viewer to immerse themselves in a fantastical world beyond imagination. 16:9

Stable Diffusion 3.5 Large

076d623bdb381c56c42ae74e17743e23.jpeg

Midjourney 6.1

d02be92d2127fab85f1b04977de76e1f.png

Flux 1.1 Pro

243aa473e85db476d3d448aff6356192.jpeg

Промпт: A realistic image of an ancient temple overgrown with lush vines and vibrant flowers, nestled deep within a dense jungle, where shafts of sunlight break through the thick canopy above to reveal intricate carvings on the stone, telling stories of a bygone era, and a small, tranquil waterfall cascading nearby, adding to the quiet beauty of this hidden gem. 16:9

Stable Diffusion 3.5 Large

f77436cedcb7f40dfce6eda4e24f679e.jpeg

Midjourney 6.1

13759235e8955ee3c39839da3a5b8a0f.png

Flux 1.1 Pro

29434a687252668784eefb0a19b5c5b2.jpeg

Промпт: An elegant illustration of a wise old owl perched majestically on a gnarled branch, surrounded by a flurry of falling autumn leaves in various shades of orange and red, with twinkling stars shimmering in the night sky above, and the quote 'Wisdom Comes With Time' artistically written beside it in an elegant script, capturing the essence of wisdom and tranquility inherent in nature. 16:9

Stable Diffusion 3.5 Large

b49ff3c07f19fd29c97e9e9c6a65488f.jpeg

Midjourney 6.1

ed2945ef9d105f23b3c24dc54575508d.png

Flux 1.1 Pro

634501ca89c4d693371beeefdc9576ee.jpeg

Выводы

Не буду растягивать вывод, так как не считаю себя способной писать адекватную и справедливую критику. Напишу коротко, тем более что самые интересные мнения мы сможем прочитать в комментариях от уважаемых посетителей Хабра.

Мне больше всего понравился Midjourney, его я выбирала чаще всего как лучший вариант. Четыре изображения вместо одного — это тоже большой плюс, и с реалистичностью изображений у него тоже все очень хорошо. Но непонятно, почему он до сих пор не научился нормально отображать слова и фразы. Хотя со словами на русском языке не справился никто из них. Flux и Stable Diffusion на первый взгляд выдают практически одинаковое качество, но при детальном сравнении их результатов, я чаще выбирала то, что создал Flux.

А что думаете вы?

© Habrahabr.ru