AudioSourceRE Sets New Standard for Real-Time, AI-Powered Audio Separation

Originally presented at the Audio Engineering Society (AES) New York convention in 2018, the release of the DeMIX Pro software established a strong new contender in the field of audio separation. Meanwhile, AudioSourceRE has used machine learning to take the concept further, while the latest version 5 of DeMIX Pro and DeMIX Essentials has set a new standard for AI-powered audio separation, making the unimaginable totally possible. Now, AudioSourceRE is working to offer the same technology, running in real-time embedded systems!
  Although many attempts already existed, audio source separation was projected to the spotlight in 2014, when Audionamix launched the ADX TRAX spectral analysis software at the 137th AES convention in Los Angeles, as reported by audioXpress. Meanwhile, in Asia, the karaoke market had already generated many sophisticated tools for those specific purposes - mainly for removing voice or specific instruments from existing tracks. But eventually, machine learning eventually proved to be extremely effective and, in Korea, Gaudio Lab’s pioneered its AI source separation as an experimental online service.

Gradually, all DJ mixing software tools started introducing AI-based extraction features and eventually played with full track separation with some degree of success. But always focusing mainly on spectral analysis, and on the possibility to manipulate and remix sounds, and not at fully isolating clean sounds of instruments, or voices, as Gaudio had already proven possible. In contrast, AudioSourceRE explored true sound isolation algorithms - the culmination of over 20 years of research into sound separation by Derry Fitzgerald (CTO). 

The AudioSourceRE technology approach proved able to offer anyone in the field of music and post-production unrivaled creative freedom. Its DeMIX Pro software effectively allowed users to isolate vocals, drums and other instrument tracks from any already-mixed audio and archive material. This proved useful for remixing, up-mixing, re-mastering, and post-production applications. DeMIX Pro proved more useful than other audio separators by offering unlimited non-destructive track separations, with convenient merge tracks functions, an advanced spectral audio editor and built-in multi-channel mixer. 
 
AudioSourceRE also released DeMIX Essentials as an entry-level tool for DJ's, producers, and remix artists. In fact, AudioSourceRE even attracted the support of veteran audio engineer, renowned educator, and source separation specialist, Rick Silva, the visionary behind the original ADX TRAX product line from Audionamix.

Fast-forward to the post-pandemic, and AudioSourceRE takes the original DSP-based, audio source separation algorithms to a whole new level, with the use of machine learning and AI. The company also hired Estefanía Cano as Chief Scientist, a research scientist that for 10 years, worked in sound classification, fingerprinting, denoising, and sound source separation algorithms at Fraunhofer IDMT in Germany.

Recently, the original AudioSourceRE technology was boosted to a whole new level, by adding the power of AI to audio detection and extraction. Today, AudioSourceRE combines AI/ML with unique perception-based algorithms that consider how the human hearing system or auditory perception works, delivering natural-sounding results that mirror the original.

More importantly, AudioSourceRE is now ready to take its technology to real-time systems and not just post-production software. Its technology is now being ported to embedded systems and to run in real time on ultra-low-power RISC-V processors, such as Greenwaves GAP9 processor, which enables applications for true wireless stereo designs such as ultra-low latency adaptive noise cancellation, and AI-powered sound elimination or enhancing.
 
AudioSourceRE Software
For now, the exciting possibilities are clearly available on the latest AudioSourceRE software tools. The latest software DeMIX Pro was used to create some excellent demonstrations. Including a video - available on Youtube - of Freddie Mercury's isolated vocals from Queen's Bohemian Rhapsody performance during Live Aid, in 1985. Not surprisingly, the sound separation demonstration video, created from the TV broadcast, was viewed well over a million times.

Also recently, AudioSourceRE launched their new VOXLESS plug-in, a real-time vocal extraction plugin for digital audio workstation (DAW) software that is able to generate high-quality separated vocal and backing tracks from any stereo music source on the fly – no cloud processing needed. The reason why these tools are so effective are the proof of the power of combining the company's digital signal processing algorithms with artificial intelligence. All that power is well-evident in the latest DeMIX Pro and DeMIX Essentials version 5 software, placing the full power of separation and track creation from previously recorded music. The refined separation workflow ensures an even better user experience, with an improved layout for efficiently handling multiple audio clips. DeMIX Pro v5 is even able to perform flawless batch processing track separation from multiple songs - a whole album at the same time if needed. Each song is divided into convenient, perfectly separated lead vocals, lead guitar, pianos, bass, drums, and everything-else. Users can even extract vocals collectively or choose any combination of lead vocals and backing vocals separately. The software makes this technology fully available to all, and the results clearly place AudioSourceRE at the leading position of this type of application.

Very soon, this technology will be used to power everyday features in personal audio devices running in real time to allows us to better understand conversations in a loud restaurant, remove annoying commentary from a live broadcast, or protect our hearing from loud noises. Those will be demonstrated at CES 2024, in Las Vegas.
www.audiosourcere.com