If you’re in any way involved in the digital audio ecosystem, the following news can’t help but get you excited as their potential to change a lot of things. TSAC is a brand-new, free and open-source, audio compression utility developed by Fabrice Bellard, the author of FFmpeg and QEMU.
Developed to compress audio at incredibly low bitrates while maintaining high perceptual quality, TSAC impressively reduces a 3.5-minute stereo song to just 192 KiB. It operates at bitrates as low as 5.5 kb/s for mono and 7.5 kb/s for stereo, all at a standard sample rate of 44.1 kHz.
However, to utilize TSAC effectively, users need to have powerful hardware, particularly any of the latest NVIDIA GPU models, which significantly speed up the compression process thanks to the CUDA technology.
While TSAC supports CPU operations, the process is notably slower and requires CPUs supporting the AVX2 instruction set.
Now, how are these exceptional levels of compression achieved? At the core of TSAC’s efficiency is a modified version of the Descript Audio Codec, adapted for stereo output and enhanced with a Transformer model.
The source code is available for download from the project page, where you can also find a number of samples for TSAC-compressed audio files. Installing it, however, requires an NVIDIA GPU with at least 4 GB of memory and CUDA version 12.x or higher.
Additionally, users must install the FFmpeg utility to convert audio files to a raw format suitable for compression. More information on this, as well as how to use the tool itself, can be found here.
