Introducing the FLUX.1 Suite
FLUX.1 is a suite of text-to-image models that define a new state-of-the-art in image detail
At BlackForest Labs, we're thrilled to unveil the FLUX.1 suite of text-to-image models, which is set to revolutionize the field of generative imaging with its superior detail, precision, and versatility. The FLUX.1 suite is available in three distinct variants, each tailored to meet specific needs:
FLUX.1 Models Overview
FLUX.1 [pro]
The pinnacle of our model lineup, FLUX.1 [pro] delivers unparalleled performance in image generation. It excels in prompt adherence, visual quality, and diversity, making it our top-tier option for high-end applications. Whether you're working on intricate projects or need top-notch image quality, FLUX.1 [pro] meets the highest standards. It is available through our API, as well as through our partners Replicate and fal.ai. For bespoke enterprise solutions, please reach out to us directly.
FLUX.1 [dev]
FLUX.1 [dev] is an open-weight model designed for non-commercial use. While it provides similar image quality and prompt adherence as FLUX.1 [pro], it offers enhanced efficiency. This variant is ideal for developers and researchers who need access to high-performance image generation without commercial constraints. FLUX.1 [dev] weights can be found on Hugging Face, and it can be tested on Replicate or Fal.ai. For commercial use, please contact us for more details.
FLUX.1 [schnell]
Our fastest model, FLUX.1 [schnell], is perfect for local development and personal use. It combines speed with efficiency, making it an excellent choice for those who need rapid image generation capabilities. FLUX.1 [schnell] is available under the Apache 2.0 license and can be accessed via Hugging Face, with inference code provided on GitHub and through Hugging Face’s Diffusers. Additionally, it integrates seamlessly with ComfyUI from day one, allowing for an easy and streamlined local development experience.
Detailed Look at FLUX.1 [schnell]
FLUX.1 [schnell] is a cutting-edge text-to-image model featuring 12 billion parameters, designed to offer a balance between speed and image quality. Here’s a closer look at what makes FLUX.1 [schnell] stand out:
High-Speed Performance: FLUX.1 [schnell] is engineered for rapid image generation, making it ideal for personal use and local development projects.
Efficient Image Generation: Trained with advanced latent adversarial diffusion distillation techniques, this model can produce high-quality images in just 1 to 4 steps.
Open Access: Released under the Apache 2.0 license, FLUX.1 [schnell] is available for personal, scientific, and commercial purposes. This open-access model encourages broad usage and integration.
Integration and Support: Users can find the model weights on Hugging Face, and detailed inference code is available on GitHub and through Hugging Face’s Diffusers. FLUX.1 [schnell] is also compatible with ComfyUI for a smooth local inference experience.
Getting Started with FLUX.1 [schnell]
For those looking to get started with FLUX.1 [schnell], follow these steps to set up and run the model using the Diffusers Python library:
1. Install or Upgrade Diffusers:
```bash
pip install -U diffusers
```
2. Run the Model:
```python
import torch
from diffusers import FluxPipeline
pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-schnell", torch_dtype=torch.bfloat16)
pipe.enable_model_cpu_offload() # Optimize VRAM usage by offloading to CPU
prompt = "A cat holding a sign that says hello world"
image = pipe(
prompt,
guidance_scale=0.0,
num_inference_steps=4,
max_sequence_length=256,
generator=torch.Generator("cpu").manual_seed(0)
).images[0]
image.save("flux-schnell.png")
```
For more detailed information, refer to the Diffusers documentation.
Limitations and Out-of-Scope Use
While FLUX.1 [schnell] offers exceptional capabilities, it’s important to be aware of its limitations and appropriate use cases:
- Factual Accuracy: This model does not provide factual information and may reflect societal biases.
- Prompt Matching: Output may not always match the given prompts precisely.
- Usage Restrictions: The model should not be used for any activities that violate laws, exploit minors, spread false information, or infringe upon individual privacy and rights. It must not be used for generating non-consensual content, illegal pornography, or disinformation campaigns.
With the FLUX.1 suite, BlackForest Labs is setting a new benchmark in text-to-image technology, providing powerful tools for a wide range of applications. Explore the potential of FLUX.1 [schnell] and other models to elevate your projects and creative endeavors.