Live media workflows are leveraging NVIDIA’s AI microservices to enhance production capabilities, but challenges arise due to network latency and bandwidth limitations when processing high-bitrate, uncompressed media streams. NVIDIA has introduced new AI reference applications on Holoscan for Media, designed for seamless integration with ST 2110 streams, allowing real-time media effects with minimal latency.
Key applications include AI virtual cameras, which utilize PyTorch and the NVIDIA DeepStream SDK to create dynamic, cropped outputs for each presenter, and automatic speech recognition (ASR), employing the Riva Parakeet NIM for real-time transcription.
To start building, developers need an AI workstation with an NVIDIA RTX Pro GPU, a Holoscan environment, and an IDE like Visual Studio Code. The latest Holoscan 25.4 update also enhances monitoring for production environments and improves automation for local setups. Unlock the potential of real-time AI for live media with Holoscan for Media today!
Source link