Flux
FreemiumState-of-the-art text-to-image models by Black Forest Labs, from open-source to production API.
A family of state-of-the-art text-to-image models from Black Forest Labs, ranging from fully open-source weights to proprietary production APIs, with advanced editing tools for inpainting, structural control, and style transfer.
What is it
Flux is a suite of AI image generation models developed by Black Forest Labs, the team behind Stable Diffusion. Built on a 12-billion-parameter rectified flow transformer with dual text encoders (CLIP + T5) and a 16-channel VAE, Flux delivers exceptional photorealism, prompt adherence, and typography accuracy. The family spans open-weight models for self-hosting and closed API tiers for production use.
What it can do
Users generate images from text prompts across multiple quality and speed tiers. Open-source variants (FLUX.1 [schnell], FLUX.2 [klein] 4B) can be run locally or self-hosted for free. API tiers (FLUX.1 [pro], FLUX 1.1 [pro], FLUX.2 [pro/max]) deliver top-tier quality via pay-per-image billing. A suite of editing tools — Fill (inpainting/outpainting), Canny (edge control), Depth (spatial control), Redux (image variation), and Kontext (multi-reference editing) — enables precise creative control.
Who is it for
It is built for AI artists demanding anatomical accuracy and legible text in generated images; developers who want to self-host open-weight models under Apache 2.0; creative agencies producing campaign visuals at scale; and studios integrating high-fidelity image generation into production pipelines.
Key Features
FLUX.1 [pro] / FLUX 1.1 [pro] / FLUX.2 [pro]
The flagship closed models delivering the highest photorealism and prompt fidelity. FLUX 1.1 [pro] generates up to 6x faster than the original pro tier, while FLUX.2 [pro] supports megapixel-based pricing for 4MP outputs. Accessible only via API.
FLUX.1 [schnell] / FLUX.2 [klein]
Fully open-source variants under Apache 2.0, free for commercial use. FLUX.1 [schnell] produces high-quality images in 1–4 inference steps. FLUX.2 [klein] 4B generates images in under one second on consumer GPUs. Both can be self-hosted at zero license cost.
FLUX.1 Fill — Inpainting & Outpainting
Edit images with precision using binary masks. Remove objects, replace elements, or expand canvases beyond original boundaries with context-aware blending that matches lighting and texture. Available in [pro] and [dev] variants.
FLUX.1 Canny — Edge-Based Structural Control
Extract Canny edges from a reference image to lock composition and object boundaries. Generate new visuals with identical structure but different styles, textures, or content. Ideal for pose replication and layout preservation.
Use Cases
Photorealistic concept art for games and film
Concept artists generate hyper-realistic characters, environments, and props with accurate anatomy and lighting. The open-weight dev models allow studios to fine-tune on proprietary visual styles.
Product photography and e-commerce visuals
Marketing teams generate studio-quality product shots on any background. FLUX 1.1 Pro Ultra delivers 4MP resolution suitable for print catalogs and large-format displays.
Fashion and apparel design prototyping
Designers sketch garments in text prompts and generate photorealistic models wearing the clothing. Redux and Kontext maintain brand consistency across seasonal lookbooks.
Architectural visualization and interior design
Architects use Depth and Canny tools to preserve floor plans and spatial layouts while iterating on materials, lighting, and furniture styles. Clients review photorealistic renders before construction begins.
Pricing plans
Frequently Asked Questions
Discussion
No comments yet. Be the first to start the thread.