Chintan Parikh, Product Manager. Shuangfeng Li arbeitet als Software-Ingenieur. Weiyi Wang arbeitet als Software-Ingenieur. Gerardo Carranza, Software Engineer. Users can enjoy seamless AI capabilities in their mobile apps, such as real-time video effects, automatic speech recognition (ASR), and motion capture. For developers, however, executing advanced models directly on devices often requires balancing distinct challenges around thermal management, battery preservation, and avoiding frame drops. LiteRT enables fast, responsive AI experiences by unlocking Neural Processing Units (NPUs)—specialized hardware designed for these workloads—without sacrificing performance. It is a cross-platform, production-ready framework for on-device AI, delivering CPU, GPU, and NPU acceleration across mobile, desktop, and IoT devices. LiteRT is engineered for performance and scalability, offering a unified API that makes it easy to deploy high-speed AI capabilities.
Google Developers Blog