FLUX.1 is a new AI model (available on Replicate) that makes images from text. Unlike most text-to-image models, which rely on diffusion, FLUX.1 uses an upgraded technique called "flow matching."
While diffusion models create images by gradually removing noise from a random starting point, flow matching takes a more direct approach, learning the precise transformations needed to map noise onto a realistic image. This difference in methodology leads to a distinct aesthetic and unique advantages in terms of speed and control.
We were curious to see how this approach impacts the generated images, so we fed it a variety of prompts, many created by other AI models. Here are some observations:
One of the challenges in text-to-image generation is accurately translating words into visual representations. FLUX.1 handles this surprisingly well, even in complex scenarios like memes.
Prompt:
Photograph of letterpress serif type on thick rough creamy paper saying 'REPLICATE.COM'
This image of letterpress type highlights how FLUX.1 can combine precise text rendering with its "flow" aesthetic. The letters are crisp, the ink looks wet. The paper is less convincing.
Prompt:
A meme of a famous actor making a funny face with the text 'When you forget your lines' in a quirky font
While it didn't quite nail a specific actor's likeness, this meme shows that FLUX.1 understands the concept. Just look at his face.
Prompt:
This is fine dog meme underwater. Text: 'Climate change is fine'
The "This is fine" dog meme, now underwater, is a perfect example of FLUX.1's ability to seamlessly blend text into an image. Well, near-seamlessly. Is fine.
FLUX.1 consistently generates high-quality images with a keen understanding of light, shadow, and texture.
Prompt:
A detailed image of a garden where the flowers are made of delicate glass, reflecting the sunlight beautifully
These glass flowers demonstrate how FLUX.1 grasps the interplay of light and material. The focus is not simply on the texture of glass, but on how light refracts and transmits through the petals, creating a luminous effect.
Prompt:
Owl feathers merging with autumn leaves in wind
FLUX.1 captures fine detail with precision. Notice how the owl feathers and autumn leaves are rendered with organic, natural textures.
FLUX.1 doesn't just imitate artistic styles; it seems to grasp their underlying principles, allowing for creative reinterpretations.
Prompt:
A cubist interpretation of a famous superhero in action
This cubist rendition of a superhero showcases FLUX.1’s ability to apply artistic principles to diverse subjects.
Prompt:
watercolor of famous wave painting
This "watercolor" version of Hokusai's Great Wave off Kanagawa offers intriguing insights into FLUX.1. Not only does it suggest the iconic wave is part of the model's training data, but it also highlights how the "flow" technique approximates the movement of pigments through water, paper, and ink.
FLUX.1 excels at composing complex scenes, placing objects and characters in a way that feels both believable and visually engaging.
Prompt:
A realistic image of an enchanted library where books float in mid-air and the shelves are made of ancient, twisted roots.
This enchanted library, with its trees growing through the bookshelves and books suspended in mid-air, showcases FLUX.1’s ability to create believable yet fantastical environments.
Prompt:
A realistic photo of a giant coffee cup being used as a hot tub by a group of friends.
FLUX.1 effortlessly captures the absurdity of a giant coffee cup hot tub. The scene is well-composed, with a clear sense of scale and playful interaction between the characters.
Perhaps the most striking aspect of FLUX.1 is its "flow" aesthetic, a consequence of the underlying flow matching technique. This gives the images a unique sense of organic movement and fluidity, almost as if the pixels themselves are in motion.
Prompt:
Dog with swirling, Van Gogh-style fur patterns
The energy in this dog's fur is almost tangible, blending directly into the whorls of paint suggestive of Starry Night.
The "flow" aesthetic is difficult to define but immediately recognizable. It evokes traditional artistic techniques like oil painting and airbrushing, imbuing the images with a dreamlike quality that sets FLUX.1 apart.
All the images in this post were generated with FLUX.1 [schnell], a version optimized for speed and local execution.
FLUX.1 [schnell] is an exciting new tool for artists, developers, and anyone interested in exploring the potential of AI image generation. Try it out on Replicate and see what you can create.