Chips&Media and Visionary.ai have unveiled what they describe as the world’s first fully AI-based Image Signal Processor (ISP), marking a significant shift away from decades-old, hardware-bound imaging pipelines toward a software-defined future.
Announced ahead of its public debut at CES 2026, the joint solution replaces traditional, fixed-function ISP hardware with an end-to-end AI software pipeline that delivers real-time improvements in video quality while remaining tunable and updatable after deployment. The companies position the technology as a foundational building block for next-generation cameras across consumer, industrial, automotive, and edge AI applications.
For years, ISPs have been designed as rigid silicon blocks, optimized at manufacturing and largely immutable thereafter. While incremental gains have been made in performance, the underlying architecture has remained static, limiting image—quality improvements and preventing meaningful post-production upgrades.
The Chips&Media–Visionary.ai collaboration aims to change that paradigm. By moving the entire ISP pipeline into AI software powered by neural networks, the new system enables dynamic adjustment of image quality, behavior, and performance characteristics, including via over-the-air updates.
The full AI ISP builds on Visionary.ai’s established AI-based denoising technology, extending it into a complete imaging pipeline. The result, according to the companies, is consistent, high-quality video across a wide range of lighting and motion conditions, with reduced blur and noise, improved sharpness, and more accurate color reproduction—all processed in real time.
Beyond visual quality, the AI-driven approach also improves downstream computer vision tasks. Chips&Media and Visionary.ai report that object detection accuracy increases by more than 75 percent in low-light environments, alongside a 91 percent reduction in false positives. These gains are particularly relevant for edge AI systems, where imaging quality directly impacts analytics, safety, and automation outcomes.
From a hardware efficiency standpoint, Chips&Media’s custom neural processing unit, WAVE-N, plays a central role. Designed to address the traditional power, bandwidth, and latency constraints of edge devices, the architecture uses line-by-line processing within convolutional neural network computations to minimize DRAM access. This approach reduces both power consumption and latency while maintaining competitive area efficiency.
Despite supporting 16-bit floating-point precision—often associated with higher accuracy but greater silicon cost—the AI ISP maintains a compact footprint, according to the company. Chips&Media attributes this balance to its more than two decades of experience in multimedia hardware IP design.
Executives from both companies framed the announcement as a step toward fully software-defined cameras. Steve Kim, CEO of Chips&Media, said the partnership combines silicon-proven IP with advanced AI software to deliver imaging systems that are “smarter, faster, and more adaptable” than existing solutions.
Visionary.ai CEO Oren Debbi added that shifting the ISP from hardware to AI software “rewrites the rules of imaging,” enabling continuous improvement in image quality and intelligence without redesigning silicon.
The AI-based full ISP will be demonstrated live for the first time at CES 2026, Booth 61701 in Eureka Park at the Venetian Expo. OEM evaluations are expected to begin shortly after the show, as companies seek to bring software-defined imaging to mainstream camera platforms.