Flux : The Most Exciting New Model

The Exciting New Model in Three Versions

Flux has introduced an exciting new model available in three versions: FLUX.1 [pro], FLUX.1 [dev], and FLUX.1 [schnell].

The image generation quality of these models is quite remarkable, especially when compared to previous models. Notably, the distortion of hands and fingers is significantly reduced. To compare overall performance, we generated images using the same prompt featuring an Asian girl showing her fingers. These images are raw and unedited.

You can compare the results yourself at https://varoriya.com/service/flux/.

The prompt used for the three images was:

The prompt used for the three images was:Korean cute girl 18yo with blue eyes and a white blouse posing for a picture, diamond rings on her finger, hand touch her face, lovely delicate face, soft portrait, beautiful delicate face, ethereal beauty, soft portrait, beautiful portrait image, feminine beautiful face, portrait soft light, portrait soft low light, gorgeous face portrait, very pretty face, large eyes

FLUX.1

All three versions utilize the Flow Rectified Transformer architecture with 12 billion parameters, ensuring high-quality image generation and excellent adherence to prompts. The images generated are more accurate to the prompts than ever before.

Sample images and models are available on GitHub.
Compatible with Comfy UI for personal use.
Prohibited for illegal activities, child exploitation, spreading false information, harassment, or creating inappropriate content.
Accessible at https://varoriya.com/service/flux/.

FLUX.1 [schnell]

Utilizes Latent Adversarial Diffusion Distillation for fast image generation (3-8 seconds on varoriya.com).
Licensed under Apache-2.0 License.
Suitable for personal, scientific, and commercial use.

FLUX.1 [dev]

Employs Guidance Distillation for improved efficiency.
Licensed under a Non-Commercial License.
Focused on research and artistic development.

FLUX.1 [pro]