BRIA

Bria develops a visual generative AI platform-as-a-service that lets developers, product teams and enterprises build image and video generation and editing features into their own products. The company's foundation models are trained exclusively on licensed data supplied by more than 30 content partners, including Getty Images, Envato, Alamy, Freepik and Depositphotos, and a patented attribution engine links generated outputs back to the underlying dataset so that data owners are compensated — an approach the company compares to revenue sharing in music streaming.

The platform is delivered as a set of layers and applications — a Visual Engine, Media Suite, Ads & Catalog and Tailored Generation — accessible through APIs, SDKs, iFrames, MCP, and Photoshop and Figma plugins, with each capability exposed as a composable skill. Central to the current positioning is VGL (Visual Generative Language), a structured, JSON-based language the models are trained to read, intended to give explicit and repeatable control over parameters such as lighting, camera angle, composition, background and objects rather than relying on natural-language prompting. Bria also publishes open models on Hugging Face, including the FIBO family of 8B text-to-image and image-to-image models and the RMBG background-removal models, and releases research papers.

Commercial and compliance guarantees are a core part of the offer: full IP indemnification for generated content, SOC 2 Type II, ISO 27001, GDPR and EU AI Act compliance, and deployment options spanning Bria Cloud, customer-managed AWS or Azure environments, on-premise and air-gapped installations. Technical customers can access source code and model weights and use ControlNets, LoRAs and LCMs for finer control.

Founding story

Yair Adato co-founded Bria in 2020 after earning a PhD in computer science focused on computer vision at Ben-Gurion University in collaboration with Harvard University and serving as CTO of Trax Retail, where he participated in the company's growth from roughly 20 employees to over 850.

Business model

Bria sells its visual generative AI capabilities as a platform-as-a-service to enterprises and developers, providing production-grade APIs/SDKs, pre-built pipelines, source-code and model-weight access, and flexible commercial terms that can draw on customers' existing cloud commitments. Revenue generated from AI-created content is shared with data partners through the company's attribution engine.

Enterprise platform-as-a-service licensing of models, APIs/SDKs and source code, with commercial terms that can leverage customers' existing cloud commitments; data partners are paid via attribution-based revenue sharing.

Traction

The company reports more than 30 data partners and cites enterprise adoption in advertising, gaming, media and retail. Its Hugging Face organization lists 24 models, 12 spaces and 34 team members, with RMBG background-removal models recording hundreds of thousands of recent downloads and the FIBO text-to-image model over 43,000. Partnerships with Nvidia, Microsoft and AWS have been formed since the March 2024 Series A, and the company received AWS Design Partner of the Year recognition.

Latest developments

In September 2025 Bria announced a Series B extension from Bright Pixel Capital to expand its PaaS offering, extend fine-tuning and personalization tooling, and apply its attribution engine to other media such as music and text. Subsequent announcements include the launch of V-RMBG 3.0, a video background-removal model, and being named winner of "Image Generative AI Solution of the Year" in the ninth annual AI Breakthrough Awards. Recent open releases center on the FIBO model family, including Fibo-Edit and Fibo-Edit-RMBG.