GPU Audio have announced that their SDK (Software Development Kit) has now been updated to support the latest graphics cards from NVIDIA and AMD, along with another significant development. Building on the support the company received for Steven Atkinson’s NAM Module, they have revealed their second SDK module offers real-time GPU-based source separation.
Showcased at the 2025 Audio Developer Conference event in Bristol, UK, the new SDK delivers genuine real-time source separation based on the HS TasNet (Hybrid Spectrogram Time-domain Audio Separation Network) model that has been researched and investigated by L-Acoustics. GPU Audio say that their new SDK module provides a neural network adapted for real-time low-latency applications, and offers huge potential for remixing and applying noise-cancelling to audio streams in real time.
The news comes alongside the announcement that GPU Audio's Pro Audio division has been rebranded, and will now be known as Soundcore.
The latest SDK is available now for Windows and Mac users. More information about its performance, along with a link to the kit on the company’s Github page, can be found via the link below.
www.gpu.audio/newsfeed/real-time-source-separation-is-here-76
