Balck Forest recently launched FLUX. 1 a new AI Image Generator
Black Forest Labs introduced its groundbreaking text-to-image AI models, FLUX.1, on Thursday. This marks a significant advancement in artificial intelligence. FLUX.1 incorporates a unique blend of multimodal and parallel diffusion transformer blocks.
Key Points
Black Forest Labs offers three FLUX.1 variants:
FLUX.1 Pro: A high-end, commercial, closed-source model.
FLUX.1 (dev): An open-weight model for non-commercial use.
FLUX.1 (schnell): A faster version for personal and local development.
Hardware Considerations
Users with less powerful GPUs might face limitations. The open-source FLUX.1 models demand substantial resources, requiring approximately 23GB of storage and nearly 24GB of VRAM to operate efficiently. This could hinder users with smaller GPUs. A compressed, “quantized” version might be released in the future.
Accessibility
Black Forest has partnered with Fal AI to offer cloud-based generation. Additionally, users can experiment with the models for free on Replicate.com. After exhausting the daily quota, a fee of $1 is charged for generating 33 images with Flux Pro or 333 images with the faster Flux Schnell version.
Background
FLUX.1’s launch follows the June release of Stable Diffusion 3 Medium by Stability AI, which faced criticism for its inaccurate human anatomy generation.
Unlike predecessors, FLUX.1 excels at producing lifelike hands and adheres closely to prompts.
About Black Forest Labs
Founded by the creator of Stable Diffusion, Black Forest Labs is a new venture backed by investors. The company’s advisory board includes industry leaders like Michael Ovitz and Matthias Bethge.
Company Focus
Black Forest is currently focused on text-to-image generation but plans to expand into video creation. Prioritizing accessibility and transparency, the company will use FLUX.1 as a foundation for a new model. This model aims to compete with Sora, Gen-3 Alpha, and Kling by offering high-definition, rapid video creation and editing capabilities.